POL00031502
POL00031502
Deloitte PRIVATE AND CONFIDENTIAL — SUBJECT TO
@ LEGAL PRIVILEGE
‘Bramble’ — Draft Report
THIS IS A DRAFT DISCUSSION DOCUMENT AND REPRESENTS A WORK IN PROGRESS AND MAY
CONTAIN PRELIMINARY RESULTS OR CONCLUSIONS INCOMPLETE INFORMATION OR INFORMATION
WHICH IS SUBJECT TO CHANGE
31 October 2016
This report and the work connected therewith are subject to the Terms and Conditions of the engagement letter dated 09 April
2014 (amended 11 March 2016) between Post Office Limited (POL) and Deloitte LLP. The report is produced for the General
Counsel of Post Office Limited (POL), solely for the use of Post Office Limited (POL) for the purpose of assessing assurance
sources and the design of certain controls relating to the Horizon system. Its contents should not be quoted or referred to in
whole or in part without our prior written consent, except as required by law. Deloitte LLP will accept no responsibility to any
third party, as the report has not been prepared, and is not intended for any other purpose.
Contents
1 Executive Summary
2 Background
3 Scope and Approach
4 Work Performed
5 Assumptions and Limitations
Appendix
12
14
32
33
POL00031502
POL00031502
POL00031502
POL00031502
1 Executive Summary
All results highlighted within this report are provisional as our QA processes remain ongoing due to outstanding
evidence from Fujitsu and as a result our findings may change as the work performed is finalised.
1.1 Background
Post Office Limited (POL) continues to respond to allegations that the “Horizon” IT system used to record
transactions in POL branches is defective and the processes associated with it are inadequate (the “Allegations”).
The ‘Allegations’ span a period of over 15 years, some pre date 2000 and others relate to 2016. In response to the
commencement of litigation proceedings, Deloitte has been instructed to plan and execute procedures and respond
to three scope areas supporting POLs ability to understand how Horizon (HNG-X) has been operated to prevent
incorrect system operation that could have resulted in Sub-postmaster detriment.
The scope areas over which Deloitte have been requested to perform procedures are as follows:
(i) Scope Area 1 - To carry out an analysis of the relevant transaction logs for branches within the
Scheme to confirm, insofar as possible, whether any bugs in the Horizon system are revealed by the
dataset which caused discrepancies in the accounting position for any of those branches.
(ii) Scope Area 2 - To carry out a full review of the use of Balancing Transactions throughout the lifetime of
the Horizon system, insofar as possible, to independently confirm from Horizon system records the
number and circumstance of their use.
(iii) Scope Area 3 - To carry out a full review of the controls over the use and capability of authorised
Fujitsu personnel to create, amend or delete baskets within. a sealed audit store throughout the lifetime
of the Horizon system, insofar as possible.
Against each of these three scope areas the main body of this report will outline further:
(i) Background and context in relation to this engagement;
(ii) The approach Deloitte have taken to planning the procedures;
(iii) The testing procedures POL has requested Deloitte undertake in response to the planning activities;
and
(iv) Results of these testing procedures.
1.2 Summary of Results
Asummary of key controls tested and results are set out below. A full set of agreed procedures tested and
associated results has been included in Section 3 of this report. These should be reviewed in tandem with the
assumptions and limitations that have been included in Section 5.
POL00031502
POL00031502
41.2.1 Scope Area dt
Scope Area 1: To carry out an analysis of the relevant transaction logs for branches within the Scheme to confirm, insofar
as possible, whether any bugs in the Horizon system are revealed by the dataset which caused discrepancies in the
accounting position for any of those branches.
We have performed testing of key inherent system controls, together with review of some of the source code which
supports the correct operation of the system in relation to ‘bugs’ (error, flaw, failure or fault in a system that causes
it to produce an incorrect or unexpected result, or to behave in unintended ways) which may have given rise to or
contributed to the allegations under investigation. These are controls which in our scoping discussion with POL and
Fujitsu have been determined to be fundamental to protecting the integrity of transaction data within the system
The key controls identified were:
« All transactions on the Horizon Counter balance to zero — No Relevant Exceptions Noted.
e Transactions are atomically (either in entirety, or not at all) written to the Branch Database — No Relevant
Exceptions Noted.
e Digital Signature controls are applied to the Message Journal during initiation of transfer to Branch Database,
ensuring the integrity of data. - No Relevant Exceptions Noted.
* Access to mechanisms for managing the digital signatures are segregated from database administration
responsibilities (via system access rights restrictions), meaning that even if such access rights be abused the
digital signature that is included with every Counter and Kiosk transaction could not be spoofed. — Relevant
Exceptions Noted.
The exception noted was
- ‘Anumber of users have access to mechanisms for managing the digital signatures and have database
administration responsibilities and access. This raises the theoretical risk of a user ‘spoofing’ the digital
signature. It is understood that for this risk to be realised, due to time limitations and volume of work
required in order to successfully ‘spoof’ the signature, a program would have to be written.’
« Transaction Acceptance (in relation to interface file receipt for non-Counter originated interface files) is required
by Sub Postmaster's in order to be accepted into branch accounting records. - No Relevant Exceptions Noted.
e Recovery processes are in place for transactions in the event of connectivity failure. - Relevant Exceptions
Noted.
The exceptions noted were:
- For one of the transaction recovery scenarios tested (whereby a user session is automatically logged
out after a period of inactivity — 59 minutes after the session screen being locked), it was noted that
Post Office business rules are in place for Horizon to automatically commit unprocessed transactions
to the branch database tables. This would have the effect of committing any unprocessed transactions
within a basket to the branch database. However when next authenticating into Horizon, after being
automatically logged out, the user is immediately presented with a till receipt confirming that the
transactions had been committed to the branch database.’
- ‘Where a new product is created, the recovery script could theoretically be coded to do nothing,
meaning no recovery of transactions would occur in the event of connection failure - no rollbacks or
roll-forwards would happen in this case.’
The first exception could lead to an increased risk that Sub-postmasters are unaware of transactions being
posted in a power failure, although they are notified by receipt that this has occurred. The second
exception could lead to the risk of inappropriate/inaccurate resolution to a recovery situation.
The above controls were tested at a recent point in time, as they are system controls. Given this limitation the
following procedures were undertaken over change control, as changes to the system are subject to the change
control process in place over the Horizon system:
016 Deloit
POL00031502
POL00031502
e A review of sources of assurance around change control was performed and it was noted that three ISAE3402
reports were performed covering the period April-December in 2012, 2013 and 2014 by professional services
firm Ernst & Young LLP. The scope of the report was seen to include 'Fujitsu's system of IT Infrastructure
Services supporting POL's POLSAP and HNG-X applications’. Within each reports’ scope was a control
objective relating to change management, and in each report reviewed no deviations were noted against this
objective, or any related controls.
e Further it was identified through change documentation review, and discussion with Fujitsu SMEs that various
controls tested had specifically changed, either since inception of HNG-X (replacing Riposte) in 2010, or
changed during the lifespan of Riposte. Please see Appendix 5 for a full list of controls tested and a view on
whether the controls have been consistent.
In summary the major change affecting the operation of controls in relation to this scope area is the creation of
the Branch Database (BRDB) to replace individual branch databases (2010). This change fundamentally
altered the operation of many controls tested. Whilst Fujitsu have attempted to give a view on controls in
operation in the Riposte system, much of the knowledge of this system has left the business.
Whilst not causing an exception to one of the controls covered by the scope of our work the following exception
relating to General IT Controls over Horizon was noted:
- One Fujitsu user has access to both development and live environments of HNG-X, contravening
typically expected segregation between environments in a change control process.
Fujitsu stated that:
“Whilst we appreciate that there is lack of segregation of duties here for Gerald between Live and
Development, it is felt that there is a strong business need for this access for Gerald. He provides 4th
line/final line support for the audit service.and is in regular weekly contact with the Security audit team to
assist them in resolving queries with the audit service. He is the lead designer/developer and system
owner.
Additionally there are compensating controls in place such as CCTV, and the auditing (performed by
Fujitsu) we have in place (and the technical controls around not being able to change audit items for 7
years) acts as a safeguard against anyone with access trying to change anything in an unauthorised way.”
In addition to the system controls noted above, the following analytics procedures were performed to support this
scope area:
e Review of the case data available (relevant to allegations) for transactions indicating items of risk from a
system functionality perspective. The analytical procedures outlined in Appendix 6 were undertaken, and a
number of items of interest were noted, see Appendix 6a for details and summary of findings. One finding of
note is that 'there were 59 (0.0019%) session ids from a total of 3,074,830 which were out of balance based on
the transactional data received. Those 59 session ids out of balance related to 16 distinct branches from 118 in
total. The session ids out of balance were all pre system migration to HNG-x in 2010.
POL investigators have been handed this information for further investigation. In short, whilst various
characteristics were noted that could be indicative of risk within the system, further manual investigation will be
required by POL's investigators to conclude. This has been discussed with POL management during the
course of our work.
1.22 Scope Area 2
To carry out a full review of the use of Balancing Transactions throughout the lifetime of the Horizon system, insofar as
possible, to independently confirm from Horizon system records the number and circumstance of their use.
POL00031502
POL00031502
In performing our procedures against this scope area, we have worked with POL and Fujitsu to identify other
methods of posting transactions which impact a branch accounts, without knowledge of the sub-postmaster which
in the context of the allegations is the more generic risk illustrated by Balancing Transactions. This highlighted
other areas of risk, such as:
e ‘Global Users’ - being central users who can access branches remotely for support purposes. Critically such
users are not able to post transactions remotely, but only when physically in the branch
e Database and Operating System Users with sufficient privileges to post transactions directly to the database
from outside of Horizon, thereby bypassing the system controls to manage activity.
These areas have been brought into scope.
In summary across each of these areas, including Balancing Transactions, controls were noted to be operating
effectively. In particular, based on the procedures we have performed:
e Logical Access rights to these sensitive functions had been appropriately restricted. - No Relevant Exceptions
Noted.
e Any writes by the Shared Service Centre (SSC) to the Branch Database (BRDB) must be audited. The
mechanism for inserting a correction record must ensure that the auditing of that action performed must be
atomic. — No Relevant Exceptions Noted.
e Access to these mechanisms is segregated from key management responsibilities (via system access rights
restrictions), meaning that should such access rights be abused the digital signature that is included with every
Counter and Kiosk transaction could not be spoofed. — Relevant Exceptions Noted.
The exception noted was:
- ‘Anumber of users have access to mechanisms for managing the digital signatures and have database
administration responsibilities and access. This raises the theoretical risk of a user ‘spoofing’ the digital
signature. It is understood that for this risk to be realised, due to time limitations and volume of work
required in order to successfully ‘spoof’ the signature, a program would have to be written.’
e It was also noted via a control walkthrough that any Transaction Corrections created by POL Finance must be
accepted by a Postmaster at branch prior to affecting branch accounts. — No Relevant Exceptions Noted.
« Inherent system controls around Global. Users were tested, notably that Global users with a Role of ADMIN
cannot log onto any Branch other than Global (including Remote access controls to branch infrastructure (e.g.
Counter)). - No Relevant Exceptions Noted.
e SSC will have privileges of only inserting balancing / correcting transactions to relevant tables in the database.
SSC will not have any privileges to update or delete records in the database. — Relevant exception noted.
The exception noted was:
- ‘The control wording is not accurate. A small number of users are granted extended privileges which
enable them to update / delete records. However the contro! is operating in line with management's
expectations. Access to the privileged role is restricted to users explicitly authorised for this access.
User actions are audit logged, and not proactively reviewed.’
The above controls were tested at a recent point in time, as they are system controls. Given the limitations around
this the following procedures were undertaken over change control, as changes to the system are subject to the
change control process in place over the Horizon system:
e A review of sources of assurance around change control was performed and it was noted that three ISAE3402
reports were performed covering the period April-December in 2012, 2013 and 2014 by professional services
firm Ernst & Young. The scope of the report was seen to include ‘Fujitsu's system of IT Infrastructure Services
supporting POL's POLSAP and HNG-X applications’. Within each reports scope was a control objective relating
to change management, and in each report reviewed no deviations were noted against this objective, or any
related controls.
POL00031502
POL00031502
e Further it was identified through change documentation review, and discussion with Fujitsu SMEs that various
controls tested had specifically changed, either since inception of HNG-X (replacing Riposte) in 2010, or
changed during the lifespan of Riposte. Please see Appendix 5 for a full list of controls tested and a view on
whether the controls have been consistent.
In summary the major change affecting the operation of controls tested is the creation of the BRDB to replace
individual branch databases (2010). This change fundamentally altered the operation of many controls tested.
It is not known whether balancing transactions existed in Riposte, as much of the knowledge of this system has
left the business.
An exception was noted relating to a core General IT Control exception around Segregation of Duties, please see
page 4 above where this issue is described in detail.
In addition to the system controls noted above, the following analytics procedures were performed to support this
scope area:
e All available audit data over the use of Balancing Transactions was inspected (12/03/2010 — 28/05/2016) and it
was noted that only 1 ‘true’ Balancing Transaction was inserted, it did not relate to a branch involved in the
allegations, and the branch was made aware of the transaction prior to.insertion. Other uses of the tool used to
insert Balancing Transactions were noted, however they did not affect transactional data and related to the
update of a specific flag (SU) to enable continued processing.
1.2.3 Scope Area’
Scope Area 3: To carry out a full review of the controls over the user and capability of authorised Fujitsu personnel to
create, amend or delete baskets within a sealed audit store throughout the lifetime of the Horizon system, insofar as possible.
In performing our procedures against this scope area, we have worked with POL and Fujitsu to identify how
baskets of transactions flow from creation at the counter, through the sealed audit store (See Background section
for a high level overview).
Further we have tested controls over the accuracy, completeness and validity of the flow of data into the audit
store, which is used as the master data point for audit purposes. We highlight the following key controls during
scoping as being fundamental to ensuring the accuracy, completeness and validity of this data flow:
e The flow of data from counter to audit store was mapped at a detailed level (See Section 1 for high level
overview). Security controls over data at rest (when held in an intermediate location), and completeness and
accuracy controls over data in transit (transfer of data from one holding location to another) including exception
monitoring were tested. - No Relevant Exceptions Noted.
« Security controls over access to the audit servers, and audit store were tested, specifically that there are
separate roles and a clear segregation between audit server administration staff, who administer the
architecture, and Fujitsu service audit staff, who have access to retrieve data from the audit store via an audit
workstation. — No Relevant Exceptions Noted.
e Access to mechanisms for managing the digital signatures are segregated from database administration
responsibilities (via system access rights restrictions), meaning that even if such access rights be abused the
digital signature that is included with every Counter and Kiosk transaction could not be spoofed. — Relevant
Exceptions Noted.
The exception noted was:
- ‘Anumber of users have access to mechanisms for managing the digital signatures and have database
administration responsibilities and access. This raises the theoretical risk of a user ‘spoofing’ the digital
POL00031502
POL00031502
signature. It is understood that for this risk to be realised, due to time limitations and volume of work
required in order to successfully ‘spoof’ the signature, a program would have to be written.’
e The ATS (Audit Track Scheduler) collects files for sealing and records a log of its activities to the ATD (Audit
Track Database). In sealing a file the seal is generated using a MDS hash algorithm. Once a file has had a seal
calculated the file is written to Centera and details are stored in the Audit Track Seal Database. — No Relevant
Exceptions Noted.
e Audit tracks and seals are copied to the equivalent import area on the remote audit server as part of Audit
server overnight schedule. On arrival, the sealer on the remote audit server recalculates the seal value of the
imported audit track and compares it with the original value in the imported seal file. Assuming they match, the
file is then written to the remote Audit archive. If the seals do not match, the Audit track and seal file are
moved to a holding area and an event is raised. Manual investigation is necessary to investigate the cause of
the discrepancy (which could be indicative of tampering with the data in between the two Audit servers). — No
Relevant Exceptions Noted.
e Audit tracks that are gathered at one data centre are replicated to the Audit server at the remote data centre. —
No Relevant Exceptions Noted.
e As Audit tracks are retrieved from the archive, their seals are checked (by re-application of the MD5 message
digest function) to ensure that the source data has not been tampered with while it was stored in the archive.
The digital signature check is also applied at this point to ensure data integrity. - No Relevant Exceptions
Noted.
« The remote directories from which the Audit Server gathers Audit Tracks is configured so that only the Audit
Server (or an administrator who has been explicitly given permission) is able to delete files in the directory. —
No Relevant Exceptions Noted.
e All users (including administrators) of the Audit Workstation and Audit Server log onto systems using two factor
authentication in conjunction with the HNG-X Active Directory system. Each user is uniquely identifiable. - No
Relevant Exceptions Noted.
e The following operating system level events on the Audit Server are audited via the System Management event
monitoring facilities:
+ Log on/Log off (including unsuccessful log on attempts)
+ File Creation, Deletion and Modification (on selected files)
+ Modifications to system. configuration (incl. software configuration and account details)
+ System start up and shut down
+ Change of user rights
Relevant Exceptions Noted:
- ‘Review of the audit settings for the Audit Server noted that the audit policy change which relates to
change of user rights was set to log success events only, with failure not enabled.’
The above controls were tested at a recent point in time, as they are system controls. Given the limitations around
this the following procedures were undertaken over change control, as changes to the system are subject to the
change control process in place over the Horizon system:
e A review of sources of assurance around change control was performed and it was noted that three ISAE3402
reports were performed covering the period April-December in 2012, 2013 and 2014 by professional services
firm Ernst & Young. The scope of the report was seen to include ‘Fujitsu's system of IT Infrastructure Services
supporting POL's POLSAP and HNG-X applications’. Within each reports scope was a control objective relating
to change management, and in each report reviewed no deviations were noted against this objective, or any
related controls.
e Further it was identified through change documentation review, and discussion with Fujitsu SMEs that various
controls tested had specifically changed, either since inception of HNG-X (replacing Riposte) in 2010, or
POL00031502
POL00031502
changed during the lifespan of Riposte. Please see Appendix 5 for a full list of controls tested and a view on
whether the controls have been consistent.
In summary it is understood controls relating to the audit server and store have been relatively consistent
throughout the lifetime of Riposte and Horizon. It should be noted that whilst Fujitsu have attempted to give a
view on controls in operation in the Riposte system, much of the knowledge of this system has left the
business.
An exception was noted relating to a core General IT Control exception around Segregation of Duties, please see
page 4 above where this issue is described in detail.
In addition to the system controls noted above, the following procedures were performed to support this scope
area:
« The process of Journal-Sequence-Numbering (each transaction is given a unique ID of 1 greater than the
previous transaction), whereby completeness checks are performed over these JSNs, is an optional setting
within the system (which assures the completeness of messages from the counter in the audit store). Testing
supported that this control has been enabled since 2010 and.not turned off since inception in 2010.
1.3 Fundamental Limitations and Assumptions
Any procedures performed during our work against each scope area are subject to a number of assumptions and
inherent limitations.
Specifically it should be noted that controls tested/to be tested for Phase 1 relating to the system will be tested on
the system (HNG-X) operating at the time of our review, and Finance controls testing will cover controls in place at
the time of our review. It must be noted that at the time of some allegations the Legacy Horizon system was still in
use, and further there is currently a refresh of POL Finance Centre controls underway. In performing our testing we
have commented on the evidence that supports the view that the control was operating in the relevant period
where we were able to do so.
Further all analytical procedures for Phase 1 are subject to the availability of data / evidence, it is noted that while a
full transactional audit log is available for up to 8.5 years, logistical / time constraints limited the volume of data that
is able to be retrieved and interrogated. Also any controls testing is subject to the availability of evidence.
Finally our work performed for Phase 0 and proposed/tested procedures for Phase 1 are specifically limited to the
three scope areas outlined in the scope section above. Our work is focused on identifying, and performing
procedures to validate, the facts in relation to the Horizon system with regard to the three scope areas as above.
Please see Section 5 for a full list of assumptions and inherent limitations.
POL00031502
POL00031502
2 Background
The Horizon system was developed by Fujitsu and is the core operational and Electronic Point of Sales (EPOS)
platform for the Post Office network. Whilst formal benchmarking data is not available, it is considered by
interviewed stakeholders to be one of the largest computer systems in existence in terms of the number of
transactions it processes on a daily basis, and it sits at the core of a complex systems estate with multiple
interfaces with other Post Office systems as well as third party systems.
The system has been in use for over 15 years and is audited by multiple parties for statutory audit, service auditor
reporting, and accreditation purposes. Given its size and scale, and the considerable intellectual property that
Fujitsu has built within the system, in relation to this piece of work, there is a significant quantity of documentation
articulating how the various modules and features comprising the system operate. Much of this documentation has
formed the focus of our review during Phase 0 of the work.
In understanding Horizon it has been important to distinguish between features which are of relevance today, and
the time period to which that relevance applies. In particular we would highlight the migration between the system
commonly referred to as Legacy Horizon, and the online variant operated today, referred to as Horizon HNG-X.
The key difference between these two iterations of the platform is the way data is stored. In the Legacy version
data was replicated between the data centre and the branches (this system was called Riposte), whilst over the
course of 2010 a migration event occurred whereby the Riposte system was replaced by the Branch Database
model, the Branch Database being a data centre only database storing the transactional and accounting data for
the branches, with a Counter application held locally within the branch which connects to the branch database as
necessary. This change may have influenced the relevance of some of the controls in existence at the present time
and care must be taken to consider this when prioritising procedures.
The Branch Database is also key to understanding the flows of data to the Audit Store given that it acts as a hub
for all branch transactional and accounting records. The diagram below provides clarity on the high level flow of
data from transaction origination through to the Audit Store:
indicative Data Flow Overview
Counter Front end of the system, located behind the ‘counter’ in
Branches. Transactions are input here by the Postmaster.
ssk Configured the same way as the Counter, but for Kiosk
(Kiosk) outlets,
BAL Transactions are bundied into ‘Baskets’ and sent from the
Counter / Kiosk to the BAL once they are complete. All
baskets must balance to 0 (Debit = Credit). Data is then
transferred from the BAL — BRDB in real time.
BROB The Branch Database is an Oracie database and sits at the
heart of the Horizon system. It receives transactions from
the BAL and also from other sources as illustrated,
Transactions input into BRDB from sources other than the
Counter/SSK are fed back to the Counter/SSK and have to
be ‘Transaction Accepted’
‘Audit The Audit Server run a Daemon process which searches
Server ‘for new data in the BRDB. When relevant transactions are
identified they are pulled into the Audit Server from the
BROB. Data is held in the Audit Server for approximately 5
days.
Audit Store After approximately 5 days data is written from the Audit
Server to the Audit Store where itis stored semi-
permanently (currently 8.5 years of data is stored).
Transactional data is stored in a message journal, whereby
the completeness of the audit data is confirmed by JSN
I sequencing
rote Audit Upon request from POL, Fujitsu audit staff can use the
ep I Desktop Audit Desktop to query the audit store to extract specified
data. Upon extraction from Audit Store - Audit Desktop, the
integrity of the data is confirmed via the digital signature
1 and seal check,
<2) ACD is produced with the requested Audit Data.
POL00031502
POL00031502
This diagram shows most but not all of the data feeds associated with the Branch Database, but does show all of
the direct transactional feeds to the Branch Database. It demonstrates the convergence of the data flows at the
Branch database and the chain of subsequent data movements.
In considering these diverse data feeds a key concept is those which use a public key infrastructure (Counter) for
completeness and accuracy of the message journals to the Branch Database, versus those which use a
combination of interface controls (header and footer records) for completeness, combined with manual
interventions from Branch staff around the completeness of the associated data (being the data feeds external to
the Horizon infrastructure e.g. Paystation).
2.1. Potential Risks
Our view of the potential risks which are inherent in the high-level procedures requested by POL are listed below.
In creating this list of potential risks we have considered the high-level procedures themselves, our understanding
of the allegations made by the sub postmasters and our knowledge of the Horizon system through workshops with
POL and Fujitsu personnel.
The table below shows how each potential risk relates to POL's scope areas:
Requested Scope Areas
1- To carry out an analysis of the
relevant transaction logs for
branches within the Scheme to
confirm, insofar as possible,
whether any bugs in the Horizon
system are revealed by the dataset
2-To carry out a full review of the
use of Balancing Transactions
throughout the lifetime of the
Horizon system, insofar as
possible, to independently confirm
from Horizon system records the
3- To carry out a full review of the
controls over the user and
capability of authorised Fujitsu
personnel to create, amend or
delete baskets within a sealed audit
store throughout the lifetime of the
which caused discrepancies in the number and circumstance of their Horizon system, insofar as
accounting position for any ofthose use. possible.
branches.
R1 v
R2 v
R3 v
R4 v v
Key to potential risks
R1. If Horizon does not process transactions correctly and these are not identified and resolved, these
could lead to sub postmaster financial loss.
R2. If inappropriate transactions can be created centrally by POL or Fujitsu which branch staff and sub
postmasters are unaware of, this would undermine the sub postmasters’ ability to trust the transactions in
Horizon are authentic and could cause sub postmaster financial loss.
R3. If data flow to the audit store is not complete, accurate or valid, the conclusions from the investigations
by case handlers or other parties dependent on these records cannot be relied on.
R4. If once data is in the Audit Store or extracted to support case investigation it is subject to
amendment, modification or deletion, this would also reduce confidence in case handlers’ conclusions.
POL00031502
POL00031502
Controls
POL management are responsible for ensuring there is a system of internal control designed to mitigate these
potential risks and that these controls are operating effectively.
No system of internal controls can be expected to guarantee the associated potential risk has not been realised.
For example, in our experience it is not reasonable to expect any enterprise software to be free from bugs
throughout the duration of its use. However, the design of enterprise software should take into account the key
risks to the application’s ongoing security and operation. Where possible inherent system controls should be
developed to prevent these potential risks being realised. Monitoring controls may also be implemented to detect
issues so they can be resolved in a timely manner by the right people. A robust change management process
should be in place to ensure only authorised changes are made and changes are tested thoroughly prior to being
implemented.
POL00031502
POL00031502
3 Scope and Approach
3.4
We have structured our work around the three scope areas POL have asked us to review, as shown in the table
below:
Scope of Work
POL consider instructing a suitably qualified
party to carry out an analysis of the relevant
transaction logs for branches within the
Scheme to confirm, insofar as possible,
whether any bugs in the Horizon system are
revealed by the dataset which caused
discrepancies in the accounting position for any
of those branches. od
POL instruct a suitably qualified party to carry
out a full review of the use of Balancing
Transactions throughout the lifetime of the
Horizon system, insofar as possible, to
independently confirm from Horizon system
records the number and circumstance of their
__use.
POL instruct a suitably qualified party to carry
out a full review of the controls over the user
and capability of authorised Fujitsu personnel to
create, amend or delete baskets within a sealed
audit store throughout the lifetime of the
POL will instruct Deloitte to
_ determine whether such an
I analysis/review is feasible, and
_ if itis, to provide an indication
_ of the cost, time and process
that would be incurred.
POL will instruct Deloitte to
determine whether such an
analysis/review is feasible, and
if it is, to provide an indication
of the cost, time and process
_ that would be incurred.
_ POL will instruct Deloitte to
undertake this review,
throughout the lifetime of the
Horizon system, insofar as is
possible.
le.
3.2 Summary of Approach and Work Performed
The work was performed in two phases. Phase 0 was ‘Discovery’ and Phase 1 was ‘Testing’.
3.2.4 Phase 0 - Discovery
This phase of work performed constituted ‘the ‘Discovery Phase’, whereby Deloitte performed initial enquiries and
investigations across the three scope areas to identify procedures which POL could undertake for each scope area.
In performing work for Phase 0, Deloitte conducted the following procedures:
e Review of relevant technical documentation as requested and provided by Fujitsu/POL during the course of this
engagement. We have set out the documentation reviewed during the course of this work in Appendix 1.
« Workshops with Finance staff in Chesterfield on 14'" and 23% March, and 18" April 2016.
e Workshop with Fujitsu staff in Bracknell on 14‘ April 2016.
* Workshop with Case Handlers in Chesterfield on 8" April 2016
POL00031502
POL00031502
The aim of these procedures was:
« Toenhance Deloitte's previous understanding of the key concepts, processes, risks and controls associated
with the Horizon system, relevant to the three scope areas highlighted above (see 2.1).
e To identify the fundamental limitations and assumptions which will need to be made and considered by
management when deciding which procedures they wish to conduct during Phase 1 (see 1.3).
e As a result of i) and ii) above the identification of possible procedures which could be adopted by management
in order to provide assurance over the risks posed in relation to the three scope areas highlighted above (see
1.3.4). We identified three core procedure types which were then utilised during Phase 1:
- Analytics - Procedures using data tools to analyse large volumes of data for particular characteristics of
interest or the absence thereof. For example verification for a given set of case data that the SN sequence
is complete.
- Controls review and testing — Verification through walkthrough, enquiry, and subsequent evidence
gathering that controls relating to the Horizon system operate as expected or otherwise, to support in
mitigation of the associated risks. For example testing the population of Fujitsu users who can administer
the Oracle DB estate underpinning Horizon directly is appropriate.
= Substantive procedures — Direct inspection of selected samples or information for confirmation of its
qualities or characteristics of note (Analytics is an example of ‘full population’ substantive procedures). In
this instance the main substantive procedures expected will be inspection of source code to verify that the
system functions as expected.
3.2.2 Phase 1 - Testing
Deloitte conducted the following procedures:
e Onsite review and visit to Fujitsu to test controls between May 2016 and September 2016.
e Review of case data provided by POL case handlers and tested for characteristics which could illustrate the
Horizon system has not operated as expected.
e Review of relevant technical documentation as requested and provided by Fujitsu/POL during the course of this
engagement. We have set out the documentation reviewed during the course of this work in Appendix 1.
POL00031502
POL00031502
4 Work Performed
4.4 Summary of Work Performed
For each scope area we have laid out our work performed as follows:
i) Setting the Scene —- We have described in a narrative format the work we have performed, and our
understanding of the relevant subject matter.
ii) A tabular format of the procedures performed in Phase 0, and the key learnings relevant to our planning.
iii) The procedures which have been performed in Phase 1 as per POL instruction, and the findings obtained
from the performance of those procedures.
4.2 Scope Area 1
Scope Area 1; Zo carry out an analysis of the relevant transaction logs for branches within the Scheme to confirm, insofar
as possible, whether any bugs in the Horizon system are revéaled by the dataset which caused discrepancies in the
accounting position for any of those branches.
4.2.1 Work Performed, and Analysis Results
Our procedures centred on the workshops and documentation reviews highlighted in Section 2.2.1 and 2.2.2. In
addition, specific to this scope area we reviewed the case data which had been provided to us, and assessed the
feasibility of performing analytics over the available case data in order to ascertain whether evidence of the system
not operating in accordance with expectations could be identified.
Our work has highlighted a number of fundamental system controls designed to ensure the integrity of processing,
and correct functionality. Key principles/items identified include:
i) Ata holistic level, IT change control processes and procedures operate over the Horizon system, and
the related controls around testing, approval, and the overall software development lifecycle should
provide assurance over the correct operation of the system. The operational effectiveness of this
control framework has, since 2012 been assessed on a regular basis, via Service Auditor Reports
(ISAE3402 produced by EY). Further sources of assurance is provided by regular 15027001
certification and ongoing audit and attestation regime, and ongoing IT focused Internal Audit and
External Audit activity. ‘Bugs’ in the system would be more likely in an environment with inadequate
change control procedures, and the level of comfort that can be gained over such controls provides a
view on the inherent risk of such errors.
ii) There are some fundamental inherent system controls, specifically designed to support correct
processing within the system. These include:
a. Journal Sequence Numbers (JSNs) are applied to each Counter transaction within the Horizon
system. These JSNs are generated using Public Key Encryption and are used by each piece of
Counter Hardware to ‘digitally sign’ a transaction. The digital signature is passed to all latter stages
of the infrastructure including the Audit Store (and beyond). This signing process provides two
critical control points over the data captured:
i. The completeness (‘density’) of the flow of transactions for a particular Branch, meaning that
completeness of the audit trail behind transactions can be ascertained.
iii)
POL00031502
POL00031502
ii. The validity and accuracy of the transactions as any changes to a transaction after the
application of the digital signature would invalidate the signature. The Audit Store extraction
routines check for this at the point of extraction.
b. Transaction Acknowledgements — Whilst JSNs are a powerful inherent system control over the
correct origination and completeness of the Message Journals from the Counter, other feeds to the
Branch Database are not subject to this control. However as an alternative control mechanism the
interface files, which issue data to the Branch Database contain Header and Footer records which
allows Horizon to automatically check the completeness of data. In addition, Branch staff accept
these interface files into their Branch accounts via Transaction Acknowledgements, meaning these
staff are directly responsible for verification that the data being received into the Branch Database
via sources outside the Counter are valid and accurate.
c. Recovery Procedures — In acknowledging that the Horizon system is dependent upon connectivity
between a data centre, a branch, and various third parties, seven recovery processes have been
designed to combat instances when a loss of connection causes an error in the completion of
transaction processes. The recovery process used depend on the nature of the connectivity issue.
Recovery scripts designed by POL are an integral part of this process.
d. The commit of transactions to the Branch Database is all performed as one Oracle DB write action,
i.e. it is atomic in nature.
e. All transactions from the Counter are checked by Horizon to ensure they. balance to zero (double
entry principle). If the Counter attempts to write a transaction which does not balance to zero, this
should be rejected via the Counter.
f. External file feeds (i.e. for data feeds not from the Counter or Kiosks) are received by the Branch
Database and into the database by Horizon before being sent to the Audit Store. Alongside this
data flow, the raw interface files are also processed directly to the Audit Store.
Alongside the inherent system controls available for our review, there are two tranches of data
analytics work that we can perform to highlight the inherent risk of system failure or ‘bugs’:
a. Using the case data we have been provided with we can perform specific profiling tests which
support the operation. of these inherent controls or rule out the occurrence of particular risky events
from within the relevant data set.
b. The BRSS (Branch Support Database) is a copy of the main Branch Database used by Fujitsu staff
for support purposes: This database contains the most recent six months’ worth of transactional
data (the Branch database itself contains only 5 days’ worth). Using tools already available via
Fujitsu we can profile this data to look for characteristics of risk (such as recovery situations,
Balancing Transactions, transactions posted by staff not related to a Branch etc).
POL00031502
POL00031502
Summary Table of Ph
Identified relevant business processes and areas of There are a set of inherent system controls within Horizon
Carry out an analysis of the relevant
transaction logs for branches within the interest. targeting the completeness, accuracy and validity of the flow of
Scheme to confirm, insofar as possible, data from Counter and other in-branch data sources, onwards
whether any bugs in the Horizon system are Review of existing technical documentation and I to Branch Database, and ultimately the Audit Store.
revealed by the dataset which caused identification of key inherent system controls. I
discrepancies in the accounting position for I Central to these controls is the digital signature applied to each
any of those branches. Workshops with Case Handlers (POL) in order to message journal of branch transactional data sent from
understand how to interpret the case data. Counter to Branch Database and beyond.
Workshops with Systems Architects (Fujitsu) inorder Connectivity issues are managed via Recovery processes, and
to understand how to interpret the case data and 80 issues with loss of connectivity have been built into the
I
technical documentation. I design of the system from the outset, in recognition this could
be an area of potential data corruption or loss.
Awalkthrough on-screen as to how the system works.
\ A strategy for our analytic procedures is to profile the available
I case data for characteristics of interest in relation to the correct
I operation of the system.
fe and Cont
POL00031502
POL00031502
4.2.3 Phase 1 Procedures
Controls Controls
1. Validate inherent system controls around: ta. No issues noted
a) All transactions on Counter system balancing to zero.
b) Atomic write and commit controls of transactions to the Branch Database. b
c) Digital Signature controls applied to Message Journal during initiation of transfer
to Branch Database.
d) Transaction Acceptance in relation to interface file receipt for non-Counter ic. Issue noted. ‘A number of users have access to mechanisms for
originated interface files. managing the digital signatures and have database administration
e) Recovery of transactions in the event of connectivity failure. I responsibilities and access. This raises the theoretical risk of a user
‘spoofing’ the digital signature. It is understood that for this risk to be
1b. No issues noted
2. Review of existing sources of assurance around Change Control and confirmation of
relevant coverage — plus targeted testing to attempt to identify changes relevant to realised, due to time limitations and volume of work required in order to
the key controls on Horizon. successfully ‘spoof’ the signature, a program would have to be written.’
Data I 1d, No issues noted
i
3. Review case data for transactions indicating items of risk from a system functionality 1e. issue noted. ‘For one of the transaction recovery scenarios tested as part of
perspective (e.g. recovery transactions are present in the case data). See Appendix recovery scenario 6, whereby a user session is automatically logged out after a
2 and6 period activity, it was confirmed that Post Office business rules are in place for
‘Horizon to automatically commit unprocessed transactions to the branch
database tables. As part of the walkthrough testing performed, it was observed
that Horizon is configured to automatically lock a user account after 15 minutes
4. Review of population of balancing transactions (to validate population of Balancing
Transactions relative to total transaction volumes)
Substantive of inactivity, at which point the user is required to re-enter their user credentials.
After a further period of 59 minutes of inactivity, Horizon is configured to
5. Review source code on screen at Fujitsu headquarters which supports the key automatically log the user out, ending a user session and committing any
inherent control operation around: unprocessed transactions within a basket to the branch database. When next
a) All transactions on counter balancing to zero. authenticating into Horizon, after being automatically logged out, the user is
b) Atomic write and commit controls of transactions to the Branch Database.
c) Digital Signature controls applied to Message Journal during initiation of transfer immediately presented with a til receipt confirming that the transactions had
to Branch Database. been committed to the branch database. From review of the printed receipt, an
d) Transaction Acceptance in relation to interface file receipt for non-Counter enhancement point was noted in that there is scope for the till receipt to include
originated interface files. further detail to the user, highlighting that an unattended transaction had
e) Recovery of transactions in the event of connectivity failure.
LLP Private and Confidential -
POL00031502
POL00031502
automatically been committed by Horizon to provide greater visibility to Post
Masters that a recovery session had been initiated.’
2. Issue noted. See Appendix 5 for details of which controls have been subject
to change.
It was noted one user has access to both development and live environments of
HNG-X.
I
Fujitsu stated that;
‘Whilst we appreciate that there is lack of segregation of duties here for Gerald
‘between Live and Development, it is felt that there is a strong business need for
this access for Gerald. He provides 4th line/final line support for the audit
service and is in regular weekly contact with the Security audit team to assist
them in resolving queries with the audit service. He is the lead
designer/developer and system owner.
i
‘Additionally there are compensating controls in place such as CCTV, and the
auditing we have in place (and the technical controls around not being able to
change audit items for 7 years) acts as a safeguard against anyone with access
trying to change anything in an unauthorised way.”
Data
3. Review of the case data available (relevant to allegations) for transactions
indicating items of risk from a system functionality perspective. The
analytical procedures outlined in Appendix 6 were undertaken, and a
number of items of interest were noted, see Appendix 6a for details and
summary of findings. One finding of note is that there were 59 (0.0019%)
session ids from a total of 3,074,830 which were out of balance based on
the transactional data received. Those 59 session ids out of balance related
to 16 distinct branches from 118 in total. The session ids out of balance
were all pre system migration to HNG-X in 2010.
~ DRAF
@ and Confidential - Subject to Legal
POL00031502
POL00031502
POL investigators have been handed this information for further investigation. In
short, whilst various characteristics were noted that could be indicative of
risk within the system, further manual investigation will be required by
POL's investigators to conclude. This has been discussed with POL
management during the course of our work.
4. No issues noted. 1 Balancing Transaction identified (in the period where
data was available for review 12/03/2010 — 28/05/2016) which did not relate
_ to a branch involved in the allegations and was appropriately approved and
governed.
su bstantive
i
{
Sa. No issues noted
5b. No issues noted
1c. No issues noted
a
5d. No issues noted
5e Post Office have the ability to create their own APADC transactions. So they
can create a product, and a transaction and then also specify the recovery
‘script which would be initiated when any of the recovery scenarios kick in.
This could, theoretically cause an issue where a new product is created, and
the recovery script is then coded to do nothing. So if the cashier sold that
product for the customer, and then in the event of the connection going down
and the recovery process kicking in - no rollbacks or roll-forwards would happen
in this case.
Our testing has shown no evidence which would suggest this has happened,
although we have not specifically performed procedures to verify this.
e and Confidential - Subject to Legal Pi
POL00031502
POL00031502
4.3 Scope Area 2
Scope Area 2: Carry out a full review of the use of Balancing Transactions throughout the lifetime of the Horizon system,
insofar as possible, to independently confirm from Horizon system records the number and circumstance of their use.
4.3.4 Work Performed, and Analysis Results
Our procedures centred on the workshops and documentation reviews highlighted in Section 2.2.1 and 2.2.2 above.
Balancing Transactions are exceptional processes used by Fujitsu support staff to correct exceptional errors in system processing/fix issues or bugs in the recording of
data. The inherent controls around the integrity of data recording are designed to ensure that such issues manifest themselves in the data on an exceptionally rare basis,
and therefore volumes of Balancing Transactions should be inherently low (substantive procedures performed support management representation there has been only 1
true Balancing Transaction since 2010).
Balancing Transactions should not be confused with Transaction Corrections which is a more routine process, used to centrally correct issues by POL Finance staff, which
are then subject to Transaction Acknowledgement by sub postmasters prior to being accepted into a Branches accounts.
Fujitsu have advised that whilst there have been several hundred instances of Balancing Transactions used throughout the known lifecycle of the HNG-X system, only one
has been a complex usage of the functionality, to correct a bug around double writing of a transaction, immediately subsequent to the migration to Horizon HNG-X. The
remainder relate to switching a flag on Stock Units (SU are a Counter concept to allocate transactions to a particular ‘sub-branch’ area to enable users to process
transactions on that stock unit (following communications failure Stock Units occasionally become locked to editing).
Our work has highlighted a number of fundamental controls which are designed within the system to control the use of Balancing Transactions and to ensure that the use
of Balancing Transactions is recorded. Key principles/items identified include:
i) Balancing Transactions are the only transactions that do not either originate at Branch, or have to be acknowledged / accepted by branch. As such the use of
Balancing Transactions is very rare.
ii) Any writes by Fujitsu Support to BRDB must be audited (record created and stored in audit store). The mechanism for inserting a correction record must
ensure that the auditing of that action is atomic with the insert of the record.
ili) Fujitsu Support with access to post Balancing Transactions cannot amend the related audit files.
iv) Fujitsu Support will have privileges of only inserting balancing / correcting transactions to relevant tables in the database. They will not have any privileges to
update or delete records in the database.
v) There are various inherent system controls around Balancing Transactions, notably that each Balancing Transaction must only contain 1 transaction (single
SQL statement) and the balancing transaction module can only be run by limited appropriate personnel.
LLP Private
Confident
POL00031502
POL00031502
In assessing the risk posed by Balancing Transactions we have also enquired as to additional ‘privileged account’ transactions which could also be used to post
transactions centrally without the knowledge of Branch staff. These enquiries have highlighted two additional areas of consideration against this risk:
a. Global Users of the Horizon System — These are users that can log on at any HNG-X Branch, and are used for a number of purposes including global user
administration.
b. Other ‘Superusers’ — At various layers of the Horizon infrastructure there exist accounts with privileged access rights which could be used to modify or insert data
relevant to transactions at branches should they not be adequately controlled. For example a superuser account on the Oracle DB forming the nucleus of the
Branch Database could insert transactions directly onto the backend (effectively Balancing Transactions are a specialised ‘legitimised’ way of using such Oracle
access).
Anumber of key controls were noted to operate on Horizon to mitigate these broader ‘superuser’ risks:
vi) Global Users are subject to two fundamental controls reducing their risks. The first is that they cannot post transactions in a branch unless they are physically
present at that branch. The second is that the Global Admins can only create users and there is therefore a Segregation of Duties between users who can create
users, and users who can post transactions.
vii) Superuser activity is monitored via log files which are transferred to the Audit Store following aggregation by the Event Management System which collects log files
from across the Horizon estate. Regardless of this control, for transactions related to the Counter and Kiosks any attempt to insert transactions into the database
by an individual with the privileged access rights to do so, would be identifiable due to the Digital Signature process applied to Message Journals from the Counter.
To circumvent this a ‘superuser’ would require the relevant access rights to the key management infrastructure which controls the Digital Signature processes, and
therefore the segregation of duties between such infrastructure and the remaining Branch infrastructure is a key control.
Alongside the inherent system controls around balancing transactions, and the completeness and accuracy of the audit log of Balancing Transactions available for our
review, there are various data analytics procedures which can be performed:
vii) As discussed above Fujitsu highlighted that while the Balancing Transaction module has been used approximately 200 times in the past 7.5 years, only 1 of
these uses has been a ‘complex’ Balancing Transaction. Analytical procedures could be performed to validate the number and nature of Balancing
Transactions which have been performed in:
a. The Case Data available
b. The BRSS most recent 6 months data available
c. The full period of data available - (7.5 years)
Sample (or full population) testing could then be performed to validate that for all Balancing Transaction records (except the 1 known Balancing Transaction,
for which the branch was aware of) no transactional postings were made using Balancing Transactions.
POL00031502
POL00031502
Summary Table of Ph
POL instruct a suitably qualified party to carry out —_ Identified relevant business processes and areas of interest. There are a sequence of inherent system controls within Horizon
a full review of the use of Balancing Transactions which ensure Balancing Transactions have certain standard
throughout the lifetime of the Horizon system, Review of existing technical documentation and characteristics, use of them is controlled, and usage is recorded in the
insofar as possible, to independently confirm from _ identification of key inherent system controls, and support in Audit Store.
Horizon system records the number and interpreting the transactional data.
circumstance of their use. Other privileged access rights which would lead to similar risks of
Workshops with Systems Architects (Fujitsu) in order to central posting of transactions with sub postmaster knowledge, such
understand how to interpret the technical documentation and _as Global Users, and ‘superuser’ accounts on the Horizon
the availability of Audit Store data. infrastructure, are also subject to key controls, most notably the
segregation of duties between the key infrastructure for digital
signatures and the infrastructure supporting the processing of Branch
transactions. These controls have been tested at a point in time.
Awalkthrough on-screen as to how the system works.
The strategy to be adopted across our analytical procedures will be to
Investigate a sample / full population of all Balancing Transaction
records found to validate the branch was aware of their usage / no
_transactional postings were made in the balancing transaction.
fe and Cont
4.3.3 Phase 1 Procedures
POL00031502
POL00031502
Controls
1. Validate inherent system controls around Balancing Transactions (See Appendix 3
for detail of controls A — 1).
2. Validate any writes by Fujitsu support staff to BRDB must be audited. The
mechanism for inserting a correction record must ensure that the auditing of that
action performed is atomic.
3. Validate Fujitsu support staff cannot amend audit files for Balancing Transactions.
4. Validate Fujitsu support staff only have privileges for only inserting balancing /
correcting transactions to relevant tables in the database. Confirm SSC do not have
any privileges to update or delete records in the database.
5. Validate broader population of Balancing Transaction controls identified. (See
Appendix 3a for detail of controls A — N)
6. Validate there is a Segregation of Duties between BRDB Administration and Key
Management Software Administration.
7. Validate inherent system controls around Global Users, notably that Global users
with a Role of ADMIN cannot log onto to any Branch other than Global (Including
Remote access controls to branch infrastructure (e.g. Counter)).
Data
8. Review case data for Balancing Transactions to validate population of Balancing
Transactions relative to total transaction volumes (Balancing transactions should be
inherently rare, and only deployed in response to actual loss/bugs in code.)
9. Review full population (already extracted by Fujitsu - 7.5 years) of balancing
transactions (sample vs full population depending on feasibility) to validate the
branch was aware of their usage / no transactional postings were made in the
balancing transaction.
e and Confidential -
Controls
No issues noted
No issues noted
No issues noted
Fe N
‘Through discussion with Fujitsu management it was noted that the
control wording is not accurate. A small number of users are granted
extended privileges which enable them to update / delete records.
However in mitigation this access is appropriately restricted to
authorised users. Users do not have the ability to bypass this role
restriction by running SUDO command. User actions are audit logged
and not proactively reviewed, and all instances of users being granted
the APPSUPP role are also captured in audit logs.’
5. Issues noted for control 2A and 2C.
2a finding noted — ‘Through discussion with Fujitsu management it was
noted that the control wording is not accurate. A small number of users are
granted extended privileges which enable them to update / delete records.
However in mitigation this access is appropriately restricted to authorised
users. Users do not have the ability to bypass this role restriction by running
SUDO command. User actions are audit logged and not proactively
reviewed, and all instances of users being granted the APPSUPP role are
also captured in audit logs.’
2c finding noted — ‘The technical document <DESAPPLLD0142> is
inaccurate. The user OPS$SUPPORTTOOLUSER does require update
access to the table BRDB_BRANCH_INFO, however the document does not
reflect this.’ This is a documentation finding only.
6. Issue noted: ‘A number of users have access to mechanisms for
managing the digital signatures and have database administration
POL00031502
POL00031502
Substantive responsibilities and access. This raises the theoretical risk of a user
I . . ‘spoofing’ the digital signature. It is understood that for this risk to be
10. Review source code on screen at Fujitsu headquarters which supports the key realised, due to time limitations and volume of work required in order to
inherent control operation around Balancing Transactions. successfully ‘spoof’ the signature, a program would have to be written.’
11. Review of Transaction Correction source code on screen at Fujitsu headquarters to
validate that Transaction Corrections must be accepted by Branches, in order to 7. No issues noted
validate Balancing Transactions are the only transactions Branches would not have
to accept. pale
12. Review the 9 Balancing Transaction Templates to validate balancing transactions 8. Review of the case data available (relevant to allegations) for
would, if the template was followed, logically perform as expected. transactions indicating items of risk from a system functionality
13. Walkthrough a Transaction Correction being raised by SCC, and the notification / Perspective. The analytical procedures outlined in Appendix 6 were
acceptance of it by a branch. undertaken, and a number of items of interest were noted, see Appendix
6a for details and summary of findings. One finding of note is that ‘there
were 59 (0.0019%) session ids from a total of 3,074,830 which were out
‘of balance based on the transactional data received. Those 59 session
ids out of balance related to 16 distinct branches from 118 in total. The
session ids out of balance were all pre system migration to HNG-X in
2010.
POL investigators have been handed this information for further
investigation. In short, whilst various characteristics were noted that
could be indicative of risk within the system, further manual investigation
will be required by POL’s investigators to conclude. This has been
discussed with POL management during the course of our work.
9. No issues noted. 1 Balancing Transaction identified (in the period where
data was available for review 12/03/2010 — 28/05/2016) which did not
relate to a branch involved in the allegations and was appropriately
approved and governed.
Substantive
10. No issues noted
11, No issues noted
e and Confidential - Subject to L DRAFT 24
POL00031502
POL00031502
12. No issues noted
13. No issues noted
© 2016 Deloitte LLP Private and Confidential - Subject to Legal Privilege - DRAFT 28
POL00031502
POL00031502
4.4 Scope Area 3
Scope Area 3: Carry out a full review of the controls over the user and capability of authorised Fujitsu personnel to create,
amend or delete baskets within a sealed audit store throughout the lifetime of the Horizon system, insofar as possible.
444 Work Performed, and Analysis Resuits
Our procedures centred on the workshops and documentation reviews highlighted in Section 2.2.1 and 2.2.2
above.
For this specific scope area our procedures centred on understanding the specific controls and processes around
protecting the integrity of data from inception to Branch Database, and subsequently to the Audit Store. Our work
highlighted a number of core concepts relevant to understanding the related risks and controls during this data
flow:
In essence the data journey can be divided into a number of distinct phases:
a. Transaction initiation within either the Counter, Kiosk, or ‘third party interface source’, and
subsequent interface to the Branch Database.
b. Archival from the Branch Database to the Audit Server.
c. Sealing of Audit Tracks via MD5 Message Digest and Archive to the Audit Store itself (Now based
on Eternis technology).
d. Subsequent Retrieval of Tracks, validation via the ARQ (Audit Track Retrieval) process, and
Investigator validation on the received data.
e. Non-Branch Transaction Data Records of Relevance
A. Transaction Initiation within either the Counter, Kiosk or ‘third party interface source’
i) For Counter and SSK (Kiosk) initiated transaction data, the SN remains a core element of control for
the Audit Store process as it validates the origination and completeness of data for a particular Counter
and is independent of the MD5 message digest elements.
ii) Given the wealth of ‘data.at rest' (stored in a directory/database awaiting onward processing) and ‘data
in transit’, security controls over access to ‘data at rest’ and interface controls over monitoring
completeness and accuracy of ‘data in transit’ are both pertinent. However the JSN concept provides
assurance regardless given interruptions in the sequence, or mis-match between signature value and
message content, would highlight downstream risks of data corruption.
iii) The other interfaces pertinent to our understanding have been represented by Fujitsu systems
architects to be:
a. Logistic Feeder Service
b. Post and Go (discontinued in 2015, but relevant prior to that date)
c. Near Real Time (NRT) feeds
d. Paystation
e. Camelot
iv) For non-Counter and Kiosk interfaces to the Branch Database completeness is provided by the
interface file header and footer record, with accuracy and validity provided by manual inspection by
Branch staff themselves via the Transaction Acknowledgements process.
v) For many of these interfaces the Post Office Data Gateway (PODG) provides the point of entry to POL
infrastructure.
LP Priva
POL00031502
POL00031502
B. Archival from the Branch Database to the Audit Server
i) Archival from the Branch Database of data take place to the Audit Server (which is the gateway to the
Audit Store infrastructure) in accordance to an automated routine which is central to the operation of
the Horizon system. If archival did not take place then very quickly the system would run out of
available capacity. Two intermediate directories are used to hold records prior to transfer to the Audit
Server.
ii) As referenced above both ‘data at rest’ and ‘data in transit’ controls are therefore relevant to this stage
of the process.
C. Sealing of Audit Tracks via MD5 Message Digest and Archive to the Audit Store itself
i) The Audit Track Gatherer (ATG) is a routine which is permanently scanning for new Audit files on the
upstream infrastructure (including the Branch Database) which. are then copied to the Audit Server,
sealed by the Audit Track Sealer (ATS), using the MD5 message digest algorithm, copied to the Audit
Store Eternis architecture itself, and then purged from the Audit Server when copied across.
ii) The Audit Server maintains a database of sealed files and their seal values, for later interrogation when
locating files, and validating their integrity has not been violated.
ili) Therefore once again both ‘data at rest’ and ‘data in transit’ controls are relevant to this stage of the
process.
iv) Once on the Eternis hardware which has now replaced the EMC Centera hardware solution, the data is
subject to a number of controls around access, deletion and amendment, all of which are designed to
maintain the integrity of the audit trail during storage. Both EMC Centera (historical solution) and
Eternis (current solution) are specialised hardware solutions for the storage of audit trail data intended
to be used forensically.
v) Previously there was a seven year limit to. the retention of data in the Audit Store, after which it was
purged by the system in line with Retention requirements. Given recent history this policy has recently
been changed to indefinite retention of all Audit Store data. As a result all transactions should be
available for as long as the Audit Store continues to exist from 04/10/2007, and therefore a complete
audit trail of all transactions ever posted on Horizon HNG-X should exist (given the migration date).
D. Subsequent Retrieval of Tracks, validation via the ARQ (Audit Track Retrieval) process, and Investigator
validation on the received data itself
i) Extraction of the data from the Audit Store is via a defined process known as the ARQ process. A
specialised Audit Desktop estate is utilised to interrogate the Audit Server database, retrieve relevant
sealed files, process the data, and burn to CD (or email as a data file), whereby it is made available to
POL investigative staff.
ii) There are a number of logical access controls operating over this process, including role based access
mechanisms, a strict ‘segregation of duties’ from POL staff and audit logs over the process.
iii) Upon receipt of the data files POL investigators carry out a number of additional checks themselves in
order to validate the data integrity.
E. Non-Branch Transaction Data Records of Relevance
i) Alongside the Branch Database data flowing into the Audit Store there are a number of other relevant
data sources:
ii) Interface files received from third party systems which are then processed into the Branch database,
are also sent directly to the Audit Store as raw files, allowing potential future reconciliation between the
two data sources.
ili) The Event Management System captures System Audit Logs from across the Horizon estate, and
processes these to the Audit Store.
POL00031502
POL00031502
Given the above understanding of the process gained from our work to date, our approach to assurance against
this scope area is largely based upon controls assurance, in combination with some limited analytics procedures to
support completeness, security and integrity of the data throughout the relevant data flows.
© 2016 Delc
POL00031502
POL00031502
Carry out a full review of the controls over the user
and capability of authorised Fujitsu personnel to
create, amend or delete baskets within a sealed
audit store throughout the lifetime of the Horizon
system, insofar as possible.
Identified relevant business processes and areas of interest.
Review of existing technical documentation and identification
of key inherent system controls, and support in interpreting
the transactional data.
Workshops with Systems Architects (Fujitsu) in order to
understand technical documentation.
Awalkthrough on-screen as to how the system works.
Walkthrough of Audit Store specific controls in order to
determine relevance and accuracy for inclusion within the
scope of our work.
The Branch Database is a key point in the data journey at which all
Branch relevant data whether generated by the Counter or by a third
party data source external to Horizon will interface to.
There are a number of intermediate points at which data is at rest
during the flow of data to the Audit Store, and understanding the
Security controls over such data will support the integrity of data
flowing into the Audit Store.
Regardless of the opportunity or otherwise for interception and
tampering of data pre its arrival in the Audit Store, for key data
originating from the Counter and the Kiosks, the digital signatures
should highlight any tampering with data prior to its usage within the
Cases.
The Case data provided can be reviewed with a view to re-performing
the key integrity checks performed by investigators, over the
completeness and accuracy of the data.
The Audit Store controls should have remained relatively constant
over the period of allegations when considering those relating to
infrastructure downstream of the Branch Database. This is due to the
HNG-X project which has influenced a number of other key control
areas, leaving the Audit Store architecture relatively untouched
44.3 Phase 1 Procedures
POL00031502
POL00031502
Controls
1. Validate Audit Store controls identified (See Appendix 4 for detail of controls 1A—
10).
2. Digital Signature controls applied to Message Journal during initiation of transfer to
Branch Database.
3. Additional Audit Store Controls identified (See Appendix 4a for detail of controls 3A.
— 3F).
4. Identification of Audit Store Data Flows at a Detailed Level, including security
controls over data at rest, and completeness, accuracy and validity controls over
data in transit.
Data
N/A
Substantive
5. Review source code on screen at Fujitsu headquarters which supports the key
inherent control operation around digitally signing transactions posted from the
Counter to the Branch Database.
6. Identification of changes relevant to the Audit Store from review of historical
documentation, and validation that the Audit Store has remained broadly consistent
over time from a controls perspective for the period relevant to the allegations.
e and Confidential - Subject to Legal ~ DRAFT
Controls
Data
N/A
No issues noted
Issue noted: ‘A number of users have access to mechanisms for
managing the digital signatures and have database administration
responsibilities and access. This raises the theoretical risk of a user
‘spoofing’ the digital signature. It is understood that for this risk to be
realised, due to time limitations and volume of work required in order to
successfully ‘spoof’ the signature, a program would have to be written.’
No Issues Noted except for control 3A.
3A finding - ‘Review of the audit settings for the Audit Server noted that
the audit policy change which relates to change of user rights was set to
log success events only, with failure not enabled.’
No issues noted
Substantive
5. No issues noted
6. See Appendix 5 for details of which controls have been subject to change.
POL00031502
POL00031502
© 2016 Deloitte LLP Private and Confidential - Subject to Legal Privilege - DRAFT 34
POL00031502
POL00031502
SA ti d Limitati
5.1. Assumptions and Limitations
Our work has been subject to the following exclusions:
1. We have not verified or tested any information or assertions provided directly by you, or directly or
indirectly by third parties;
2. For scope areas 1, 2 and 3, only matters relating to Horizon Features and Audit Store within the
Horizon processing environment have been considered during our workshops and discussions;
3. We have not provided a legal or any other opinion as to the completeness and accuracy of processing
of Horizon at any point throughout the work;
4. We have not had direct contact with any third parties other than named contacts that you have
provided to us (Appendix 1);
5. We have not reviewed any contractual provisions in place between you and third parties;
6. Our work was limited by gaps existing in the information available, relating to both the granularity of
information and the existence of the Horizon Features’ over the entire timeline of operation of Horizon
and suspense account process documentation. The effect of which is that there are in gaps within what
we are able to comment upon over this timeline;
7. We have not validated or commented on the quality of the Assurance Work? supplied to us.
Our work was also based on the assumption that the documents provided and assertions made are a complete and
accurate representation of the Horizon design, audit store process and suspense account process. We therefore
cannot comment as to whether other processes would need consideration in the context of the Matters.
We have performed work on control in place and operating at the time of the review, and not those operating at the
time of the allegations. Other evidence has been obtained, where available, to provide evidence as to whether the
control was likely to have operated at the time of the allegations.
1 “Horizon Features” is a term we have introduced to represent those features of the Horizon processing environment, including IT management
and business use controls, which provide that
* Movements in Branch ledgers have the full ownership and visibility of sub-postmasters; and
* Audit trails kept by the system are complete and accurate
2 Since its implementation in branches, POL has commissioned or has received a number of pieces of work relating to the Horizon processing
environment, to provide comfort over its integrity. This work, referred to in our report as the “Assurance Work’, provides documented assertions
relating to aspects of the design and operation of the Horizon processing environment. The Assurance Work includes IT project documents;
operational policies and procedures; internal and external investigations and reviews; independent audits; and emails confirming otherwise
verbal assertions.
016 Deigitie LLP Priv
POL00031502
POL00031502
.
Appendix I
Documents Reviewed
Document Ref Document Title
DES/APP/HLD/0047 HNG-X Counter Application High Level Design
DES/APP/HLD/0020 _ Branch Database High Level Design
DES/APP/HLD/0030 Audit Data Collection and Storage High Level Design
_DES/APP/HLD/0029 - Audit Data Retrieval High Level Design
ARC/SOL/ARC/0006 _HNG-X Architecture - Global Users
__DEV/APP/LLD/0065 __BRDBC002 — BRDB Message Journal Auditing LLD
__DEV/APP/LLD/0014 _Host Branch Database Audit Archive Purge Low Level Design
__DEV/APP/LLD/0142 Host BRDB Transaction Correction Tool Low Level Design
__DES/APP/SPG/0001 __ Host branch database support guide
DEV/APP/LLD/0199 _ Schema definition for branch database, standby branch database and branch
. support system i
_DES/APP/HLD/0035 __Exceptions and logging frameworks high level design.
DES/APP/IFS/0002 _ HNG-X:RDDS to Branch Database - Counters and HBS Reference Data and
_ Memo Submission Interface Specification
_ DES/APP/IFS/0012 L e Specificatio
_DES/APP/HLD/0083 __HNG-X Counter Subsystem : Recovery Management
_DES/APP/HLD/0021 4 Branch Database Scheduling High Level Design :
__DES/APP/IFS/0007 __ Branch Database to Legacy Host Interface Specification I
__DES/APP/IFS/0001 _.HNG-X: RDMC / RDDS to Branch Database Application Interface Specification
_ DES/APP/HLD/0049 _,,HNG-X Generic Reports Data Extract HLD I
_DES/APP/HLD/0057 _HNG-X Counter Infrastructure: Service and Process Control High Level Design __
ARC/SOL/ARC/0001 __HNG-X Solution Architecture Outline
Aud Retrieval Low Level Design
s ' POL Archiving Strategy
DEV/INF/ION/0001 Archive Server Configuration
DES/SEC/HLD/0003. HNG-X KEY MANAGEMENT HIGH LEVEL DESIGN
DES/APP/HLD0041 _HNG-X Counter Applications: Business Logic Subsystem High Level Design
_ DES/APP/IFS/0018 ____XML Message Audit between Counter or HBS and BAL/OSR
DES/APP/HLD/0012 __. DVLA Internal Web Service High Level Design
ARC/SEC/ARC/0003 il IG-X Technical Security Architecture
__DEV/APP/LLD/0204 __Host BRDB Update Outstanding Recovery Transaction Tool Low Level Design _
__DES/APP/HLD/0070 / Host Applications Monitoring High Level Design
DEV/APP/LLD/0151 _ HNGX BRDB HOST: BRANCH SUPPORT DATABASE LOW LEVEL DESIGN
POL00031502
POL00031502
Individuals Interviewed
Name Job Title
Patrick Bourke POL - ‘Bramble’ Project Manager
Mark Underwood POL - ‘Bramble’ Project Manager
Rodric Williams POL — POL Legal
Rod Ismay POL - Head of Finance Service Centre
Lorraine Garvey POL - Enquiries Manager
Sarah Haywood POL - Finance Team Leader
Tracy Middleton POL - Finance Team Leader
Michael Harvey Fujitsu -Head of Commercial
Pete Newsome Fujitsu - Business Change Manager
Torstein Godeseth Fujitsu - Chief Archite
Steve Bansal Fujitsu - Senior Service Delivery Manager
Alan Holmes Fujitsu - Customer Solution Architect
Gerald Barnes Fujitsu -Senior Software and Solutions Designer
Gareth Seemungal
ior Software and Solutions Designer
Appendix 2
Scope area I — Potential Analytics Procedures
Ref I Analytics Procedure
A Completeness Test - Identify gaps in audit log sequencing
o
Completeness Test - Identify gaps in transaction times during working hours
Cc Completeness Test - Identify two user logon events in sequence without the expected logoff event in
between, an indicator of a connectivity issue
D Completeness Test - Identify recovery transactions
E Accuracy Test - Identify zero valued transactions
F Accuracy Test - Identify branches which are out of balance based on transactional data available (should
not be possible based on inherent system controls).
G Integrity Test - Identify transactions posted by non-branch users without subsequent branch
acknowledgement.
H Integrity Test - Identify balancing transactions.
6 Dei
POL00031502
POL00031502
Appendix 3
Scope area 2 — Balancing Transactions Controls
Ref Control Description
A SSC will have privileges of only inserting balancing / correcting transactions to relevant tables in the
database. SSC will not have any privileges to update or delete records in the database.
B If the process fails (e.g. transaction file is found to be invalid), then the transaction file will not be moved
and an error message will be written to standard output.
Cc Any writes by the SSC to BRDB must be audited. The mechanism for inserting a correction record must
ensure that the auditing of that action performed must be atomic,
Appendix 3a
Scope area 2 — Balancing Transactions Controls (Broader population)
Ref Control Description
A All inserts will be audited in the table BRDB_TXN_CORR_TOOL_ JOURNAL.
B The PL/SQL package PKG_BRDB_TXN_CORRECTION will be owned by Oracle user
“OPS$SUPPORTTOOLUSER’
c The PL/SQL package PKG_BRDB_TXN_CORRECTION will execute with the permissions of the
OPS$SUPPORTTOOLUSER account and can only insert rows into the transaction tables as controlled by
an entry in BRDB_SYSTEM.PARAMETERS. The account will not have update or delete privileges.
D Each of the transaction tables that are allowed to have balancing transactions inserted on them has an
associated template file. Each file contains a template of an INSERT statement for that table, in the
required format, and listing all of the columns on the table. Users should create their own transaction file
based upon the relevant template file, substituting the values they require into the SQL. Note that some of
the column values specified in the template should not be changed — these are annotated with comments
as appropriate.
E When execution is complete the file is then moved to directory ‘/app/brdb/trans/support/brdbx015/output’
and the log file is created in directory ‘/app/brdb/trans/support/brdbx015/log’. Log file will be named using
the following convention:
<transaction_file_name>_<CCYYMMDDHHMISS>.log
Access to these 2 directories is appropriately restricted.
F It is expected that only a small number of skilled staff will run this tool and that they will have detailed
guidance as to when and how to use the tool (For example by restriction of staff to
“OPS$SUPPORTTOOLUSER’).
G From the Unix command prompt, execute the following
./BRDBX015.sh MyTransactionFile.sqI 2001
where the first parameter is the transaction file name and the second parameter is the branch code where
the balancing transaction is going to be applied. Note that the branch code must exist in the database, and
must not be for a closed branch. If this is not the case, then an error message will be shown and the run
aborted.
POL00031502
POL00031502
Ref Control Description
H The correction tool places a number of constraints on the contents of the transaction file. These are
necessary in order to provide a defined baseline upon which it can base its operation. If any of the
constraints are violated then validation will detect it and abort the run with a meaningful error message.
The constraints are as follows:
e The transaction file must be less than 32K in size
« The transaction file must only contain Unix-style end of line markers (EOL), not DOS format end of line
markers (CR/EOL)
« The transaction file can only contain a single SQL statement. If more than one balancing transaction is
required then more than one transaction file must be created, each of which is executed with a separate run
of the tool
e If the transaction file contains an introductory comment, then it must be a ‘/* ...... */ style comment, not
Aim "style comment
« The closing ‘*/' of the introductory comment must have a trailing space (i.e. ‘..... */‘)
* The run symbol at the end of the SQL must be a ‘;’ , not ‘/’, and must have a trailing space (i.e. ‘.....; ‘)
« The SQL must be a valid SQL statement according to the normal Oracle SQL parsing rules (e.g. valid
syntax, objects accessible etc)
« The SQL must begin with ‘INSERT INTO OPS$BRDB.’ and be of the form ‘INSERT INTO ..... SELECT
see FROM dual, (SELECT ..... FROM .... WHERE .....)’.
« The table name must be one of the tables named in the
BRDB_TXN_CORRECTION_ALLOWED_TABLES1 or
BRDB_TXN_CORRECTION_ALLOWED_TABLES2 configuration parameters
e All of the columns that exist on the table in question must be explicitly named. It is not necessary for
every listed column to be on a separate line, but this is advisable for readability.
« The values to be inserted must be provided by the ‘SELECT ... FROM dual ...'. Each value must be on
a separate line. Trailing comments are allowed, but must be a ‘-- .....’ style comment. Any such comment
must not include any commas. All columns must have values provided for them (even if that value is NULL).
* Certain columns are common between a subset of the transaction tables. In some cases, these columns
should be set to the same value no matter what table is in use. With the exception of the bind variables
listed earlier, the value that the SQL will try to insert is under the control of the user (i.e. it is determined by
the value specified in the SQL). However, the tool can be configured to validate that the value specified in
the sat matches that expected. In order to do this, set the
BRDB_TXN_CORRECTION_ENFORCED_VALUES configuration parameter to include the field and the
required value.
The parameter is populated as a comma-delimited list of name/value pairs, where the name is the name of
the column name, and the value is the value to be enforced. As released, this configuration parameter is
set to:
NODE_ID=99,APP_SERVER_NODE_NAME=999,BRANCH_USER=:bind_SSC_user,BRDB_INSTANCE
_NAME=:bind_instance_name
which, for example. ensures that if a ‘node_id’ column exists on the transaction table, it's value is specified
as 99. If there is no ‘node_id’ on the transaction table, then no value is enforced for that field. Note that if
the parameter does not exist, then no values are enforced in the SQL.
I The SQL statement being executed will be logged in the table BRDB_TXN_CORR_JOURNAL. The format
of the data to be written to the column JOURNAL_XML is:
“<?xml version="1.0" encoding="UTF-8"?>
<Support_Insert>
<Unix_User>Unix User Name</Unix_User>
<Oracle_User>Oracle User Name</Oracle_User>
<Sql>SQL Statement</SqI>
</Support_Insert>”
where :
e Unix User Name is the Unix user name under which the user logged in
POL00031502
POL00031502
Ref Control Description
e Oracle User Name is Oracle user that is carrying out the actual insert i.e. SUPPORTTOOLUSER
e SQL Statement is the final (i.e. after substituting actual values for bind variables) SQL that is executed
to insert the balancing transaction
J As records are being written to the audit files, the process must optionally be able to monitor if the set of
Journal-Sequence-Numbers for a node in a Branch is dense. The check should only be performed when
the value of mandatory System-Parameter ‘JOURNAL_SEQ_DENSE_SET_CHECK_ENABLED' is
“TRUE”. When a missing journal entry is encountered, a message should be written on standard output
along the lines of “...records between sequence numbers M and N are missing...”. Once the list of
auditable messages for a node is completed, an Operational exception should be raised to indicate the
count of missing sequence numbers. Duplicate records are not possible due to the primary key on this
table.
K Unix shell script BRDBX015.sh which is in the /app/brdb/trans/support/brdbx015 directory. It is deliberately
kept separate from the standard $BRDB_SH directory so that access to the script and the associated
components can be restricted to authorised users. The shell script calls the PL/SQL package
PKG_BRDB_TXN_CORRECTION.
L PL/SQL package PKG_BRDB_TXN_CORRECTION, which resides within the Branch Database and is
owned by Oracle user OPS$SUPPORTTOOLUSER. The PL/SQL package is the component that validates,
creates and audits the balancing transaction.
M If an Oracle node/instance failure occurs, the utility will fail with an error code of 99. For all other failures, it
will fail with an error code of 1 and log an operational exception in BRDB_OPERATIONAL_EXCEPTIONS.
N The SQL in the transaction file is validated as follows. Any validation failures are displayed to standard
output and logged to the log file.
e Check that the file does not contain any carriage returns, indicating DOS format EOL markers
* Check that the SQL in the transaction file parses according to the standard Oracle rules (e.g. syntax,
privileges etc). This is done using the standard Oracle DBMS_SQL.PARSE procedure.
* Check that there is only a single SQL statement in the transaction file. Note that in most cases, this will
be detected by the previous parsing step. However, the fact that the parsing does this is not described in
the Oracle documentation, so it may be changed in future releases of Oracle. Therefore, this validation
provides security if the behaviour of the Oracle procedure is changed at a later date.
« Check that the SQL begins with INSERT INTO OPS$BRDB."
* Check that the table named in the SQL is one of the tables listed in the two
BRDB_TXN_CORRECTION_ALLOWED_TABLES<n> configuration parameters. Note that as long as the
privileges are set up correctly (i.e. OPSS$SUPPORTTOOLUSER only has insert privileges on the allowed
tables), any attempt to insert a balancing transaction on a non-allowed table will cause the previous parsing
step to fail (because the user would not have the necessary privileges). Therefore, this validation provides
security in case the privileges are not correctly set up.
* Check that all the columns named in the SQL exist on the table, and that all the columns on the table
are named in the SQL
e Check that the values to be inserted are provided by a SELECT ... FROM dual, (SELECT ... FROM ...
WHERE) i.e. not a VALUES
e Check that if any of the name/value pairs that are listed in the
BRDB_TXN_CORRECTION_ENFORCED_VALUES configuration parameter are present on the table, they
are set to the listed value.
fe} Balancing transaction audit files (BRDBC033), unlike the files produced by BRDBC002, are not
compressed, but are still encrypted.
POL00031502
POL00031502
Appendix 4
Scope area 3 — Audit Store Controls Listing
Ref i Control Description
A Audit tracks that are gathered at one data centre are replicated to the Audit server at the remote data
centre. This replication process is managed by the Audit Track Sealer. As Audit tracks are secured to the
Audit archive, they are moved to an export area awaiting transfer to the remote campus. A second file,
containing the calculated seal value for the audit track is also stored in the export area.
B Audit tracks and seals are copied, using robocopy, to the equivalent import area on the remote audit server
as part of Audit server overnight schedule. On arrival, the sealer on the remote audit server recalculates the
seal value of the imported audit track and compares it with the original value in the imported seal file.
Assuming they match, the file is then written to the remote Audit archive. If the seals do not match, the
Audit track and seal file are moved to a holding area and an event is raised. Manual investigation is
_ Necessary to investigate the cause of the discrepancy.
c There will be a single instance of the ATS that concurrently accepts files for sealing/seal checking from ATG
and ATR and notifies sealed files to the ATD and into the Sealer Database for subsequent use by the Audit
Track Extractor.
The ATS shall collect files for sealing via I-ATS-4 and shall write a log of its activities to the ATD via I-ATS-
2. In sealing a file the seal shall be generated using a secure hash algorithm, the MDS algorithm has been
selected.
Once a file has had a seal calculated the file will be written to Centera and details will be stored in the Audit
I Track Seal Database via I-ATS-5.
D Access to the Audit Track files for gathering shall be via Samba (for Unix systems) or NTFS (for Windows
systems). Access to the sub directory shall be limited to.the application generating the Audit Track and the
Audit Track Gatherer. Audit track files should be written in write-append mode.
E All users (including administrators) of the Audit Workstation and Audit Server shall log onto systems using
two factor authentication in conjunction with the HNG-X Active Directory system. Each user shall be
__uniquely identifiable. CN ie as
F The remote directories from which the Audit Server gathers Audit Tracks will be configured so that only the
Audit Server (or an administrator who has been.explicitly given permission) is able to delete files in the
i directory.
G All Audit Server and Audit Workstation and Centera hardware shall be held in physically secure areas where
physical access to the systems is controlled.
H There shall be separate roles for:
° Audit Server (inc. Audit Workstation) Administration
. Fujitsu Services Audit Staff
The roles shall be mutually exclusive, i.e. no one individual shall be given access rights of more than one
role.
J The Fujitsu Services Audit Staff role shall not have any write, modify or delete access to the Audit Archive.
J The following integrity checks will be applied to the data
. Completeness of data — contiguous message sequence numbers
° Integrity of individual messages
° For Riposte data the message CRC should be checked
° For HNG-X data the message signature will be verified
Separate Riposte and HNG-X summaries of the results of the integrity checks are generated. They should
detail:
. Summary of the message sequence runs broken down by counter Id. This should include start and
end date/times and start and end message sequence numbers. Any gaps in the message sequence runs
must be highlighted.
<2 Summary of messages that have failed individual message integrity checks
POL00031502
POL00031502
Ref Control Description
Any failure of the data integrity checks will not prevent subsequent execution of the query. The audit
_workstation user will be warned of the failure via the server process status notification mechanism. -
K As Audit tracks are retrieved from the archive, they are seal checked (by re-application of the MD5 message
digest function) to ensure that the source data has not been tampered with while it was stored in the
_ archive.
L Only authorised users may access the Audit workstation applications. Authorised users are required to log
on to the workstation using two factor authentication and the HNG-X Identity Management system. An
Active Directory group named AUDIT_USER will be created with the rights required to utilise the workstation
i applications. Authorised users will be added to this group.
M All retrievals of audit data are performed using the Audit Extractor Client, and all such user actions are
L themselves audited. It is not possible for users to access the archive by any other means.
N Audit workstations and Atalla NSPs are located in secure areas. Only authorised users are given physical
__access to these areas. .
fe} All auditable messages logged during a calendar day will be made available to the audit system in
uncompressed form as a part of Branch Database batch overnight processing.
The message journal is implemented in the form of a single Oracle table named
BRDB_RX_MESSAGE_JOURNAL. Uniqueness is controlled at the level of a Branch counter using a dense
sequence known as the Journal-Sequence-Number
Appendix 4a
Scope area 3 — Audit Store Controls Listing (broader population)
Ref I Control Description
A The following operating system level events on the Audit Server will be audited via the System Management
event monitoring facilities:
« Log on/Log off (including unsuccessful log on attempts)
« File Creation, Deletion and Modification (on selected files)
e Modifications to system, configuration (inc software configuration and account details)
e System start up and shut down
e Recovery actions
e Exception conditions
e Change of user rights
B The Audit Server Administrator role shall have full access to manage all of the Audit Server and Audit
Workstation file stores and shall be granted the necessary Windows privileges.
Cc POL staff will not be given direct access to the Audit Workstation to safeguard other parts of the HNG-X
system. Instead nominated Fujitsu Services personnel will supply audit information as requested by Post
Office.
D User Log/On events are included in the Windows event log. Users are allocated to a specific role which
enables them to access the Audit databases.
E Baskets are stored for a defined period of time. The configuration of this parameter and the audit trail
around changes to it need to inspected in order to provide assurance over the maintenance time period for
audit purposes.
Appendix 5
Change Control — list of controls and their change dates.
Scope I Control
Area Ref.
1 ta
1 1b
1 tc
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Control/Procedure description
All transactions on counter
must balance to zero.
All controls of transactions to
the branch database are
atomically written and
committed.
A Digital Signature is applied
to Message Journal during
initiation of transfer to Branch
Database.
No
Details of
the change
(inc. change
reference)
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
Yes
POL00031502
POL00031502
If Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied in Riposte.
In Riposte this control is of less
importance given each Branch
operated its own database.
There is no visibility of an
reconciliation controls in place
between local and central
databases in Riposte.
Digital Signature did not exist in
Riposte. However a CRC check
was applied, which whilst Fujitsu
assert that this is less complex
than the digital signature check,
and it is noted that this check
has not been tested in detail, if
operating correctly the check
would notify Fujitsu on retrieval
of audit data from the audit store
if any amendments to data had
been made.
Controt
Ref.
Scope
Area
1 1d
1 te
1 3
Control/Procedure description
Any non-Counter originated
interface files (POLSAP or
third party sources) must be
Transaction Accepted by the
Branch.
In the event of connectivity
failure there is a transaction
recovery process which is
initiated.
Review case data for
transactions indicating items
of risk from a system
functionality perspective (e.g.
recovery transactions are
present in the case data).
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
R13 and
Yes R13.05
No 4
N/A Data N/A Data
Procedure = Procedure
Appropriately
approved and
tested?
Release
notes
obtained and
reviewed.
Seen to
document
various
management
reviews /
approvals
and testing
‘steps.
N/A Data
Procedure
POL00031502
POL00031502
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
Fujitsu assertion on
whether control has
changed since HNG-X
if Yes - detail of process in place
before change
The changes
introduced are
assumed to be 'Win in
Mails'. As part of this
initiative an extra file is
received from
Paystation and used to
trigger Track and
Trace messages (to
Royal Mail). Items on
hand are updated
reflecting postal items
delivered to and from
the branch but there is
no financial impact on
the branch from this.
The transactions
impacting the financial
state of the branch are
received in the same
file as previously - i.e.
via Transaction
Acceptance.
N/A - see
change to left N/A - see change to left
As each branch operated its
own database, transaction
. Yes recovery processes were of less
importance in Riposte.
N/A Data Procedure NIA Data N/A Data Procedure
Procedure
Scope
Area
Controt
Ref.
Sa
Control/Procedure description
Review source code on
screen at Fujitsu
headquarters which supports
the key inherent control
operation around digitally
signing transactions posted
from the Counter to the
Branch Database.
Review of existing sources of
assurance around Change
Control and confirmation of
relevant coverage — plus
targeted testing to attempt to
identify changes relevant to
the key controls on Horizon.
Review of population of
balancing transactions (to
validate population of
Balancing Transactions
relative to total transaction
volumes)
Review source code on
screen at Fujitsu
headquarters which supports
the key inherent control
operation around:
Refer to control 1.1
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
NIA (this
procedure)
N/A Data
Procedure
No
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
NIA (this NIA (this
procedure) procedure) NA (this procedure)
N/A Data N/A Data
Procedure Procedure NIA Data Procedure
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
Yes
N/A (this
procedure)
N/A Data
Procedure
POL00031502
POL00031502
if Yes - detail of process in place
before change
Source code was reviewed at a
point in time. The Digital
Signature did not exist in
Riposte. However a CRC check
was applied, which whilst Fujitsu
assert that this is less complex
than the digital signature check,
and it is noted that this check
has not been tested in detail, if
operating correctly the check
would notify Fujitsu on retrieval
of audit data from the audit store
if any amendments to data had
been made.
N/A (this procedure)
N/A Data Procedure
Source code was reviewed at a
point in time. Please refer to 1.1-
1.5.
POL00031502
POL00031502
Evidence
reviewed
Controt indicates George Sppropriately I Fulitey cesertion or Maing if Yes - detail of process in place
Ref. Control/Procedure description sald (inc, change oo and bahia peal ie Fujitsu/ Deloitte I before change
Since une. I reterence) 9 knowledge?
X (2010)?
1 Sb Refer to control 1.2
1 Se Refer to control 1.3
4 5d Refer to control 1.4
1 Se Refer to control 1.5
Any writes by Fujitsu support
staff to BRDB must be
audited. The mechanism for
2 2 inserting a correction record No - - - N/A
must ensure that the auditing
of that action performed must
be atomic.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
Fujitsu support staff cannot
2 3 amend audit files for No : : * N/A
Balancing Transactions.
Fujitsu support staff will have
privileges of only inserting
balancing / correcting It is not known whether
2 4 transactions to relevant No _ . . N/A Balancing Transactions (or
tables in the database. SSC equivalent) and associated tool
will not have any privileges to existed in Riposte.
update or delete records in
the database.
Review case data for
Balancing Transactions to
validate population of
Balancing Transactions N/A Data N/A Data N/A Data N/A Data
relative o total transaction Procedure Procedure = Procedure NIA Data Procedure Procedure NIA Data Procedure
volumes (Balancing
transactions should be
inherently rare, and only
Scope I Control
Area
Ref.
10
Control/Procedure description
deployed in response to
actual loss/bugs in code.)
Review source code on
screen at Fujitsu
headquarters which supports
the key inherent control
operation around Balancing
Transactions.
Validation there is a
Segregation of Duties
between BRDB
Administration and Key
Management Software
Administration.
Validate inherent system
control around Global Users,
that Global users with a Role
of ADMIN cannot log onto to
any Branch other than Global
(Including Remote access
controls to branch
infrastructure (e.g. Counter)).
Review a sample of the full
population (already extracted
by Fujitsu - 7.5 years) of
balancing transactions to
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
No
N/A Data
Procedure
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
N/A Data
Procedure
N/A Data
Procedure
Fujitsu assertion on
whether control has
changed since HNG-X
N/A Data Procedure
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
N/A
No
Yes
N/A Data
Procedure
POL00031502
POL00031502
if Yes - detail of process in place
before change
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
The Digital Signature did not
exist in Riposte. However a
CRC check was applied, which
whilst Fujitsu assert that this is
less complex than the digital
signature check, and it is noted
that this check has not been
tested in detail, if operating
correctly the check would notify
Fujitsu on retrieval of audit data
from the audit store if any
amendments to data had been
made.
Fujitsu represented that no such
equivalent role or ability to
remote access onto counters
existed in Riposte.
N/A Data Procedure
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
validate the branch was
aware of their usage / no
transactional postings were
made in the balancing
transaction.
Review of Transaction
Correction source code on
screen at Fujitsu
headquarters to validate that
2 rr] Transaction Corrections must No . . . N/A Source code reviewed at a point
be accepted by Branches, in in time.
order to validate Balancing
Transactions are the only
transactions Branches would
not have to accept.
Review the 9 Balancing
Transaction Templates to
validate balancing
2 12 transactions would, if the No ‘ : - N/A
template was followed,
logically perform as
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
expected.
Release
notes
obtained and
reviewed. The mechanisms for
Walkthrough of a Transaction Seen to producing TAs
Correction being raised b: Release document changed at Release
2 13 SCC, and the notification 1 Yes 55 various 55 ae aresult of See Left See Left
acceptance of it by a branch. management _ introducing Client File
reviews / Delivery.
approvals
and testing
steps.
Controt
Ref.
ta
Sa
5b
Sc
Control/Procedure description
SSC will have privileges of
only inserting balancing /
correcting transactions to
relevant tables in the
database. SSC will not have
any privileges to update or
delete records in the
database.
All inserts will be audited in
the table
BRDB_TXN_CORR_TOOL_
JOURNAL.
The PL/SQL package
PKG_BRDB_TXN_CORREC.
TION will be owned by
Oracle user
“OPS$SUPPORTTOOLUSE
R’.
The PL/SQL package
PKG_BRDB_TXN_CORREC
TION will execute with the
permissions of the
OPS$SUPPORTTOOLUSER
account and can only insert
rows into the transaction
tables as controlled by an
entry in
BRDB_SYSTEM_PARAMET
ERS. The account will not
have update or delete
privileges.
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
No
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
Fujitsu assertion on
whether control has
changed since HNG-X
- N/A
- N/A
- N/A
- N/A
POL00031502
POL00031502
if Yes - detail of process in place
before change
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
Scope
Area
Controt
Ref.
5d
Se
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Control/Procedure description
Each of the transaction
tables that are allowed to
have balancing transactions
inserted on them has an
associated template file.
Each file contains a template
of an INSERT statement for
that table, in the required
format, and listing all of the
columns on the table. Users
should create their own No
transaction file based upon
the relevant template file,
substituting the values they
require into the SQL. Note
that some of the column
values specified in the
template should not be
changed — these are
annotated with comments as.
appropriate.
When execution is complete
the file is then moved to
directory
‘fapp/brdb/trans/support/brdb
x015/output' and the log file
is created in directory
‘fapp/brdb/trans/support/brdb No
x015/log’. Log file will be
named using the following
convention:
<transaction_file_name>_<C
CYYMMDDHHMISS>.log
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
N/A
N/A
POL00031502
POL00031502
if Yes - detail of process in place
before change
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
Details of
the change
{inc. change
Pre HNG-X
change to
Fujitsu / Deloitte
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
if Yes - detail of process in place
Area Ref. before change
2
2 tb
2 5f
2
2
59
2
Access to these 2 directories
is appropriately restricted
If the process fails (e.g.
transaction file is found to be
invalid), then the transaction
file will not be moved and an
error message will be written
to standard output.
It is expected that only a
small number of skilled staff
will run this tool and that they
will have detailed guidance
as to when and how to use
the tool.
From the Unix command
prompt, execute the following
./BRDBX015.sh
MyTransactionFile.sq! 2001
where the first parameter is
the transaction file name and
the second parameter is the
branch code where the
balancing transaction is
going to be applied. Note that
the branch code must exist in
the database, and must not
be for a closed branch. If this
is not the case, then an error
message will be shown and
the run aborted
since HNG- I Teference)
X (2010)?
No - -
No - -
No - -
knowledge?
N/A
N/A
N/A
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
POL00031502
POL00031502
The SQL statement being
executed will be logged in
the table
BRDB_TXN_CORR_JOURN
AL. The format of the data to
be written to the column
JOURNAL_XML i
“<?xml version
encoding="UTF-8"?
<Support_Insert>
<Unix_User>Unix User
Name</Unix_User>
<Oracle_User>Oracle User
Name</Oracle_User>
<Sql>SQL Statement</Sql>
2 5i </Support_Insert>” No - - - N/A
where :
+ Unix User Name is the Unix
user name under which the
user logged in
+ Oracle User Name is
Oracle user that is carrying
out the actual insert i.e.
SUPPORTTOOLUSER
+ SQL Statement is the final
(i.e. after substituting actual.
values for bind variables)
SQL that is executed to
insert the balancing
transaction
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
e and Confidential -
Scope
Area
Controt
Ref.
1c
5j
Control/Procedure description
Any writes by the SSC to
BRDB must be audited. The
mechanism for inserting a
correction record must
ensure that the auditing of
that action performed must
be atomic. There also needs
a level of obfuscation to
ensure that the audit
mechanism is robust.
As records are being written
to the audit files, the process
must optionally be able to
monitor if the set of Journal-
Sequence-Numbers for a
node in a Branch is dense.
The check should only be
performed when the value of
mandatory System-
Parameter
‘JOURNAL_SEQ_DENSE_S
ET_CHECK_ENABLED'’ is
“TRUE”. When a missing
journal entry is encountered,
a message should be written
on standard output along the
lines of “...records between
sequence numbers M and N
are missing...”. Once the list
of auditable messages for a
node is completed, an
Operational exception should
be raised to indicate the
count of missing sequence
numbers. Duplicate records
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
No - es -
No - - -
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
As each branch operated its
own database, BRDB did not
exist in Riposte.
JSN check in its current format
did not exist in Riposte.
However Fujitsu assert that a
data density check was applied.
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
5k
SI
5m
are not possible due to the
primary key on this table.
Unix shell script
BRDBX015.sh which is in the
/app/brdb/trans/support/brdb
x015 directory. It is
deliberately kept separate
from the standard
$BRDB_SH directory so that
access to the script and the
associated components can
be restricted to authorised
users. The shell script calls
the PL/SQL package
PKG_BRDB_TXN_CORREC
TION.
PL/SQL package
PKG_BRDB_TXN_CORREC
TION, which resides within
the Branch Database and is
owned by Oracle user
OPS$SUPPORTTOOLUSER
. The PL/SQL package is the
component that validates,
creates and audits the
balancing transaction.
No
No
If an Oracle node/instance
failure occurs, the utility will
fail with an error code of 99.
For all other failures, it will No
fail with an error code of 1
and log an operational
exception in
N/A
N/A
N/A
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
It is not known whether
Balancing Transactions (or
equivalent) and associated tool
existed in Riposte.
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Details of
the change
{inc. change
reference)
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
BRDB_OPERATIONAL_EXC
EPTIONS.
The SQL in the transaction
file is validated as follows.
Any validation failures are
displayed to standard output
and logged to the log file.
+ Check that the file does not
contain any carriage returns,
indicating DOS format EOL
markers
+ Check that the SQL in the
transaction file parses
according to the standard
Oracle rules (e.g. syntax,
privileges etc.). This is done Itis not known whether
2 5n using the standard Oracle No r . . N/A Balancing Transactions (or
DBMS_SQL.PARSE equivalent) and associated tool
procedure. existed in Riposte.
+ Check that there is only a
single SQL statement in the
transaction file. Note that in
most cases, this will be
detected by the previous
parsing step. However, the
fact that the parsing does this
is not described in the Oracle
documentation, so it may be
changed in future releases of
Oracle. Therefore, this
validation provides security if
the behaviour of the Oracle
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
procedure is changed at a
later date.
+ Check that the SQL begins
with ‘INSERT INTO
OPS$BRDB..
+ Check that the table named
in the SQL is one of the
tables listed in the two
BRDB_TXN_CORRECTION
_ALLOWED_TABLES<n>
configuration parameters.
Note that as long as the
privileges are set up correctly
(i.e.
OPS$SUPPORTTOOLUSER
only has insert privileges on
the allowed tables), any
attempt to insert a balancing
transaction on a non-allowed
table will cause the previous
parsing step to fail (because
the user would not have the
necessary privileges).
Therefore, this validation
provides security in case the
privileges are not correctly
set up.
+ Check that all the columns
named in the SQL exist on
the table, and that all the
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
columns on the table are
named in the SQL.
+ Check that the values to be
inserted are provided by a
SELECT ... FROM dual,
(SELECT ... FROM ...
WHERE) i.e. not a VALUES
* Check that if any of the
name/value pairs that are
listed in the
BRDB_TXN_CORRECTION
_ENFORCED_VALUES
configuration parameter are
present on the table, they are
set to the listed value.
Balancing transaction audit
files (BRDBCO33), unlike the It is not known whether
files produced by Re . " . NA Balancing Transactions (or
BRDBC002, are not equivalent) and associated tool
compressed, but are still existed in Riposte.
encrypted.
POL00031502
POL00031502
Evidence
reviewed
indicates svi i
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
Control/Procedure description control has
changed
since HNG-
X (2010)?
The correction tool places a
number of constraints on the
contents of the transaction
file. These are necessary in
order to provide a defined
baseline upon which it can
base its operation. If any of
the constraints are violated
then validation will detect it
and abort the run with a
meaningful error message.
The constraints are as
follows:
+ The transaction file must be
less than 32K in size
+ The transaction file must
only contain Unix-style end of It is not known whether
2 5h line markers (EOL), not DOS No . . . N/A Balancing Transactions (or
format end of line markers equivalent) and associated tool
(CR/EOL) existed in Riposte.
+ The transaction file can only
contain a single SQL
statement. If more than one
balancing transaction is
required then more than one
transaction file must be
created, each of which is
executed with a separate run
of the tool
+ If the transaction file
contains an introductory
comment, then it must be a
‘* ...... 7 style comment, not
a‘-......’ Style comment
+ The closing ‘*/’ of the
introductory comment must
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
have a trailing space (i.e.
ST)
+ The run symbol at the end
of the SQL must be a ‘;’, not
‘/, and must have a trailing
space (i.e. af
+ The SQL must be a valid
SQL statement according to
the normal Oracle SQL
parsing rules (e.g. valid
syntax, objects accessible
etc.)
+ The SQL must begin with
‘INSERT INTO OPS$BRDB.”
and be of the form ‘INSERT
INTO ..... SELECT .....
FROM dual, (SELECT ....
FROM .... WHERE .....)’.
+ The table name must be
one of the tables named in
the
BRDB_TXN_CORRECTION
_ALLOWED_TABLES1 or
BRDB_TXN_CORRECTION
_ALLOWED_TABLES2
configuration parameters
+ All of the columns that exist
on the table in question must
be explicitly named. It is not
necessary for every listed
column to be on a separate
line, but this is advisable for
readability.
+ The values to be inserted
must be provided by the
‘SELECT ... FROM dual ...’.
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
Each value must be on a
separate line. Trailing
comments are allowed, but
must be a ‘-- ..... ‘style
comment. Any such
comment must not include
any commas. All columns
must have values provided
for them (even if that value is
NULL).
+ Certain columns are
common between a subset of
the transaction tables. In
some cases, these columns
should be set to the same
value no matter what table is
in use. With the exception of
the bind variables listed
earlier, the value that the
SQL will try to insert is under
the control of the user (i.e. it
is determined by the value
specified in the SQL).
However, the tool can be
configured to validate that
the value specified in the
SQL matches that expected.
In order to do this, set the
BRDB_TXN_CORRECTION
_ENFORCED_VALUES
configuration parameter to
include the field and the
required value.
The parameter is populated
as a comma-delimited list of
name/value pairs, where the
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Scope I Control
Area Ref.
name is the name of the
column name, and the value
is the value to be enforced.
As released, this
configuration parameter is
set to:
NODE_!ID=99,APP_SERVER
_NODE_NAME=999,BRANC.
H_USER=:bind_SSC_user,B
RDB_INSTANCE_NAME=:bi
nd_instance_name
which, for example. ensures
that if a ‘node_id’ column
exists on the transaction
table, it's value is specified
as 99. If there is no ‘node_id’
on the transaction table, then
no value is enforced for that
field. Note that if the
parameter does not exist,
then no values are enforced
in the SQL.
Scope
Area
Controt
Ref.
ta
Control/Procedure description
Validate inherent system
controls around Global
Users, notably that Global
users with a Role of ADMIN
cannot log onto to any
Branch other than Global
(Including Remote access
controls to branch
infrastructure (e.g. Counter)).
Audit tracks that are
gathered at one data centre
are replicated to the Audit
server at the remote data
centre. This replication
process is managed by the
Audit Track Sealer. As Audit
tracks are secured to the
Audit archive, they are
moved to an export area
awaiting transfer to the
remote campus. A second
file, containing the calculated
seal value for the audit track
is also stored in the export
area.
Digital Signature controls
applied to Message Journal
during initiation of transfer to
Branch Database.
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
No - -
No - -
Fujitsu assertion on
whether control has
changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
Yes
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Fujitsu represented that no such
equivalent role or ability to
remote access onto counters
existed in Riposte.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Digital Signature did not exist in
Riposte. However a CRC check
was applied, which whilst Fujitsu
assert that this is less complex
than the digital signature check,
and it is noted that this check
has not been tested in detail, if
operating correctly the check
would notify Fujitsu on retrieval
of audit data from the audit store
Scope I Control
Area
Ref.
Control/Procedure description
Identification of Audit Store
Data Flows at a Detailed
Level, including security
controls over data at rest,
and completeness, accuracy
and validity controls over
data in transit.
Review source code on
screen at Fujitsu
headquarters which supports
the key inherent control
operation around digitally
signing transactions posted
from the Counter to the
Branch Database.
Identification of changes
relevant to the Audit Store
from review of historical
documentation, and
validation that the Audit Store
has remained broadly
consistent over time from a
controls perspective for the
period relevant to the
allegations.
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
No
Yes
Details of
the change
{inc. change
reference) I ested?
Release
notes
obtained and
R10.20 reviewed.
(Refresh of Seen to
Eternis document
Storage various
infrastructur. management
e) reviews /
approvals
and testing
steps.
Appropriately
approved and
Fujitsu assertion on
whether control has
changed since HNG-X
Agree that the system
changed to the extent
that it is now
implemented on
different hardware. A
crucial point is that the
audit data was not
changed and the
digital signatures.
created in the
branches at the time
that transactions were
carried out were
persisted and
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
Yes
N/A - see
change to left
POL00031502
POL00031502
if Yes - detail of process in place
before change
if any amendments to data had
been made.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Source code reviewed at a point
in time. Digital signature check
in its current form originated in
HNG-X
N/A - see change to left
Scope
Area
Controt
Ref.
tb
1c
Control/Procedure description
Audit tracks and seals are
copied, using robocopy, to
the equivalent import area on
the remote audit server as
part of Audit server overnight
schedule. On arrival, the
sealer on the remote audit
server recalculates the seal
value of the imported audit
track and compares it with
the original value in the
imported seal file. Assuming
they match, the file is then
written to the remote Audit
archive. If the seals do not
match, the Audit track and
seal file are moved to a
holding area and an event is
raised. Manual investigation
is necessary to investigate
the cause of the discrepancy.
There will be a single
instance of the ATS that
concurrently accepts files for
sealing/seal checking from
ATG and ATR and notifies
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
No
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
demonstrate that the
data in the audit trail
has not been
tampered with.
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Scope
Area
Controt
Ref.
1d
Control/Procedure description
sealed files to the ATD and
into the Sealer Database for
subsequent use by the Audit
Track Extractor.
The ATS shall collect files for
sealing via I-ATS-4 and shall
write a log of its activities to
the ATD via I-ATS-2. In
sealing a file the seal shall be
generated using a secure
hash algorithm, the MD5
algorithm has been selected.
Once a file has had a seal
calculated the file will be
written to Centera and details
will be stored in the Audit
Track Seal Database via I-
ATS-5.
Access to the Audit Track
files for gathering shall be via
Samba (for Unix systems) or
NTFS (for Windows
systems). Access to the sub
directory shall be limited to
the application generating
the Audit Track and the Audit
Track Gatherer. Audit track
files should be written in
write-append mode.
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
Details of
the change
{inc. change
reference)
Appropriately I Fujitsu assertion on
approved and I whether control has
tested?
changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Scope
Area
Controt
Ref.
te
3a
Control/Procedure description
All users (including
administrators) of the Audit
Workstation and Audit Server
shall log onto systems using
two factor authentication in
conjunction with the HNG-X
Active Directory system.
Each user shall be uniquely
identifiable.
The following operating
system level events on the
Audit Server will be audited
via the System Management
event monitoring facilities:
+ Log on/Log off (including
unsuccessful log on
attempts)
+ File Creation, Deletion and
Modification (on selected
files)
+ Modifications to system
configuration (Inc. software
configuration and account
details)
+ System start up and shut
down
+ Recovery actions
+ Exception conditions
+ Change of user rights
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
No - - - No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Scope
Area
Controt
Ref.
tf
1g
th
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Control/Procedure description
The remote directories from
which the Audit Server
gathers Audit Tracks will be
configured so that only the
Audit Server (or an No
administrator who has been
explicitly given permission) is
able to delete files in the
directory.
All Audit Server and Audit
Workstation and Centera
hardware shall be held in
physically secure areas
where physical access to the
systems is controlled.
Yes
There shall be separate roles
for:
+ Audit Server (Inc. Audit
Workstation) Administration
+ Fujitsu Services Audit Staff
The roles shall be mutually
exclusive, i.e. no one
individual shall be given
access rights of more than
one role.
The Fujitsu Services Audit
Staff role shall not have any
write, modify or delete
access to the Audit Archive.
No
Details of
the change
{inc. change
reference)
R10.10 and
R10.20
(Refresh of
Eternis
Storage
infrastructur
e)
Appropriately
approved and
tested?
Release
notes
obtained and
reviewed.
Seen to
document
various
management
reviews /
approvals
and testing
steps.
Fujitsu assertion on
whether control has
changed since HNG-X
Agree that the system
changed to the extent
that itis now
implemented on
different hardware.
Operational processes
were not changed.
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
POL00031502
POL00031502
Evidence
reviewed
indicates
Control/Procedure description control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Pre HNG-X
change to if Yes - detail of process in place
Fujitsu/ Deloitte I before change
knowledge?
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Control
Ref.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
1 No No applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
The Audit Server Whilst it has not been
Administrator role shall have corroborated by review of
full access to manage all of technical documentation /
3b the Audit Server and Audit No No testing it is expected this control
Workstation file stores and applied pre HNG-X.
shall be granted the Fujitsu attested that controls
necessary Windows surrounding the audit store have
privileges. remained largely unchanged.
POL staff will not be given Whilst it has not been
direct access to the Audit corroborated by review of
Workstation to safeguard technical documentation /
3c other parts of the HNG-X No No testing it is expected this control
system. Instead nominated applied pre HNG-X.
Fujitsu Services personnel Fujitsu attested that controls
will supply audit information surrounding the audit store have
as requested by Post Office. remained largely unchanged.
The following integrity checks . .
will be applied to the data:
Whilst it has not been
corroborated by review of
4j + Completeness of data — No technical documentation /
contiguous message No testing it is expected this control
sequence numbers
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Scope
Area
Controt
Ref.
Control/Procedure description
+ Integrity of individual
messages
0 For Riposte data the
message CRC should be
checked
o For HNG-X data the
message signature will be
verified
Separate Riposte and HNG-
X summaries of the results of
the integrity checks are
generated. They should
detail:
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
Yes
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
For Riposte CRC control above
was in place.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Scope I Control
Area Ref.
Control/Procedure description
+ Summary of the message
sequence runs broken down
by counter Id. This should
include start and end
date/times and start and end
message sequence
numbers. Any gaps in the
message sequence runs
must be highlighted.
+ Summary of messages that
have failed individual
message integrity checks
Any failure of the data
integrity checks will not
prevent subsequent
execution of the query. The
audit workstation user will be
warned of the failure via the
server process status
notification mechanism.
As Audit tracks are retrieved
from the archive, they are
seal checked (by re-
application of the MDS
message digest function) to
ensure that the source data
has not been tampered with
while it was stored in the
archive.
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Details of
the change
{inc. change
reference)
Appropriately I Fujitsu assertion on
approved and I whether control has
tested? changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
No
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Scope
Area
Controt
Ref.
11
3d
im
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
Control/Procedure description
Only authorised users may
access the Audit workstation
applications. Authorised
users are required to log on
to the workstation using two
factor authentication and the
HNG-X Identity Management
system. An Active Directory
group named AUDIT_USER
will be created with the rights
required to utilise the
workstation applications.
Authorised users will be
added to this group.
No
User Log/On events are
included in the Windows
event log. Users are
allocated to a specific role
which enables them to
access the Audit databases.
No
All retrievals of audit data are
performed using the Audit
Extractor Client, and all such
user actions are themselves — No
audited. It is not possible for
users to access the archive
by any other means.
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Scope I Control
Area
Ref.
in
to
3e
Control/Procedure description
Audit workstations and Atalla
NSPs are located in secure
areas. Only authorised users
are given physical access to
these areas.
All auditable messages
logged during a calendar day
will be made available to the
audit system in
uncompressed form as a part
of Branch Database batch
overnight processing.
The message journal is
implemented in the form of a
single Oracle table named
BRDB_RX_MESSAGE_JOU.
RNAL. Uniqueness is
controlled at the level of a
Branch counter using a
dense sequence known as
the Journal-Sequence-
Number
Baskets are stored for a
defined period of time. The
configuration of this
parameter and the audit trail
around changes to it need to
inspected in order to provide
assurance over the
maintenance time period for
audit purposes.
Evidence
reviewed
indicates
control has
changed
since HNG-
X (2010)?
No
No
No
Details of
the change
{inc. change
reference)
Appropriately
approved and
tested?
Fujitsu assertion on
whether control has
changed since HNG-X
Pre HNG-X
change to
Fujitsu / Deloitte
knowledge?
No
No
No
POL00031502
POL00031502
if Yes - detail of process in place
before change
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
Whilst it has not been
corroborated by review of
technical documentation /
testing it is expected this control
applied pre HNG-X.
Fujitsu attested that controls
surrounding the audit store have
remained largely unchanged.
POL00031502
POL00031502
Appendix 6
Case Data Analytics Overview
The below analytical procedures were performed on 'Case Data’. ‘Case data’ refers to transactional data provided by POL, which had been extracted by Fujitsu from the
audit store, and relates specifically to the branches involved in the ‘allegations’. The data extracted is in 1 month periods relating specifically to the period of the allegations
for each specific branch.
Relevant Analytics
Scope Area POL Instruction Proposal Analytic
Procedures
1 POL consider instructing a suitably qualified party to carry + POL will instruct Deloitte to determine Review case data for 1, 2,3, 4, 4a,
out an analysis of the whether such an analysis/review is transactions indicating 5, 6, 6a, 7
relevant transaction logs for branches within the Scheme feasible, and if it is, to provide an items of risk from a system
to confirm, indication of the cost, time and functionality perspective
insofar as possible, whether any bugs in the Horizon process that would be incurred. (e.g. recovery transactions
system are revealed are present in the case
by the dataset which caused discrepancies in the data).
accounting position for any of those branches.
Tab Index Description
Analytic 1 Identify gaps in audit log sequencing
Analytic 2 Identify gaps in transaction times during working hours
Analytic 3 Identify two user logon events in sequence without the expected logoff event in between; an indicator of a connectivity issue
Analytic 4 Identify recovery transactions
Analytic 4a Identify recovery transactions that indicate a connectivity issue
Analytic 5 Count of zero valued transactions summarised by product
Identify branches which are out of balance based on transactional data available (should not be possible based on inherent system
Analytic 6 controls).
Analytic 6 Group and Session _ Identify branches which are out of balance based on transactional data available (should not be possible based on inherent system
id controls).
Analytic 7 Identify transactions posted by non-branch users without subsequent branch acknowledgement.
Appendix 6a
Case Data Summary Findings
POL00031502
POL00031502
POL investigators have been handed this information for further investigation. In short, whilst various characteristics were noted that could be indicative of risk within the
system, further manual investigation will be required by POL’s investigators to conclude. This has been discussed with POL management during the course of our work.
Procedure
Analytic 1: Identify gaps in audit log
sequencing
Analytic 2: Identify gaps in transaction times
during working hours
Analytic 3 : Identify two user logon events in
sequence without the expected logoff event
in between, an indicator of a connectivity
issue
Analytic 4: Identify recovery transactions
Analytic 4a: Identify recovery transactions
that indicate a connectivity issue
Analytic 5: Identify zero valued transactions
Comments
In order to identify gaps in audit log sequencing, the
transactions data was sorted into ascending order on
session id and txn id, and any gaps in the sequence at
both the session and txn level were identified.
In order to identify gaps in transaction times during
working hours, the transaction data was ordered by
branch, date and time. Gaps that were significantly
higher than the average gaps in transaction times were
identified, only transactions with the same date were
compared. Transactions with a stock unit of ATM, LOT,
OOH or BUR were excluded.
In order to identify two user logon events in sequence
without the expected logoff event in between, an
indicator of a connectivity issue the events data was
ordered by date and time and logon events (event code
12 or “EPOSSTransaction.Ti of Logon Completed”) not
followed directly by a log off event (event code 13, 27
and 102 or “EPOSSTransaction.Ti of Logoff Completed”)
were identified.
In order to identify recovery transactions the
eventDetailMsg column of the Events data was searched
for words like ‘successfully recovered’ but not like 'No
recovery required.’
In order to identify connectivity issues of none recovery
transactions the eventDetailMsg column of the Events
data was searched for words like ‘could not recover’ and
'No recovery required."
In order to Identify zero valued transactions, all
transactions with a sale value of 0, a quantity not equal
to zero and a mode of either 1 or SC for 'Serve
RAFT
Summary
There were 212,372 (1.60%) gaps in audit log sequencing
from a total of 13,307,999 transactions
There were 46,528 (0.35%) gaps in transaction times that
were more than 20 times higher than the average
transaction gap of all stores with the same number of
positions from a total of 13,307,999 transactions
There were a total of 1,064 (0.93%) logon events in
sequence without the expected logoff between; from a total
of 114,491 log on/off events.
There were 30 (0.00057%) recovery transactions identified
from a total of 5,289,369 transactions in the events data
There were 258 ‘no recovery’ transactions that indicate a
connectivity issue from a total of 5,289,369 transactions in
the events data
There was a total 1,314,761 (9.88%) zero valued
transactions with a quantity not equal to zero from a total of
Procedure
Analytic 6: Identify branches which are out of
balance based on transactional data
available (should not be possible based on
inherent system controls).
Analytic 7: Identify transactions posted by
non-branch users without subsequent branch
acknowledgement.
Comments
Customer’ were identified and a summary per item is
produced.
In order to identify branches which were out of balance
based on transactional data available (which should not
be possible based on inherent system controls), the
transactions data was summarised by branch (Group)
and session id and those session ids that do not sum to
zero were identified, and are ordered by balance
descending.
In order to identify transactions posted by non-branch
users without subsequent branch acknowledgement, any
users whose id did not take the usual format (6 digits - 1s
letter of forename followed by 1% and 2" letters of
surname and numeric 001) were identified. A user id of
*PS98 are Paystation transactions.and were ignored
here, a user id beginning with a * are identified as global
users
DRAFT
POL00031502
POL00031502
Summary
13,307,999. These transactions were against a total of 814
products
There were 59 (0.0019%) session ids from a total of
3,074,830 which were out of balance based on the
transactional data received. Those 59 session ids out of
balance related to 16 distinct branches from 118 in total. The
session ids out of balance were all pre system migration to
HNG-x in 2010.
There were 17 (3.03%) users (*DSI02, *JBA03, *TAKO1,
*PJO07, *BMA01, *JCA01, *RCRO1, *DCU02, *JHOOS,
*RLY01, *DWA01, *MWE01, *STU03, *GDRO1, *NSTO1,
*PJO02 and *GMU01) from a total of 561 users classified as
non-branch users who posted transactions
POL00031502
POL00031502
Appendix 7
Clarification questions
The below clarification questions and associated answers attempt to provide clarity on queries arising from the
content of this report.
Key questions
1. From the perspective of the Group Action, we are trying to understand:
a) Whether Fujitsu can edit or delete transactions recorded by branches in a way that could impact on
the branch's overall accounting position?
Yes — Transactions can be deleted at database layer (BRDB) by DBA’s.
Before audit store access locked down, transactions could be deleted at audit store level (and still
can be once a transaction has been in the audit store for 7 years), but this would not affect a
branches overall accounting position unless there was a query that resulted in the extraction of data.
if data was extracted from the audit store and records had been tampered with or removed, this
would be flagged upon extraction by the process to report on data integrity, so it would be
transparent that the data has been edited. It should be noted the warning that the data integrity check
failed can be ignored by the operator.
b) How difficult it would be to do (a)?
Access to do (a) is restricted to appropriate personnel by Fujitsu. For users who have DBA access on
the BRDB, this could be done.
However if the edit/deiete of the transaction was not done before the data had been ‘collected’ by the
Audit Server (typically.every 15 minutes), then this would not affect the record of data in the Audit
Store. The audit store is the location where data is retrieved from in the event of a dispute.
Further if the edit/delete of the transaction was performed prior to the data being ‘collected’ by the
Audit Server, whilst it would be reflected in the audit store data, upon retrieval of branch data from
the audit store, if a transaction had been removed, the ‘data density’ check would highlight a missing
transaction. If upon retrieval of branch data from the audit store a transaction had been amended, the
digital signature check would highlight an issue with the integrity of the data.
c) . Whether (a) is possible without leaving a “footprint” that is visible to either (i) postmaster or (ii) Post
Office / FJ.
i) Amendment / deletion of transactions would not be overtly notified to the Postmaster, however if
the amendment / deletion happened at the BRDB, this would affect the declarations made by
Postmasters (encouraged to do so on a daily basis) and also declarations are required to be done in
order to rollover into the next accounting period (typically 4-5 weeks). The monthly Branch Trading
Staternent which a Postmaster must sign off on in order to roll into the next accounting period would
also be impacted by a change of this nature which would capture summarised totals of transactional
data, which could be reconciled by branch back to the granular transaction log reports. All of the
mentioned reports are mechanisms by which the Postmaster would be made aware of any such
changes.
Amendment / deletion of data in the audit store has no effect on branch accounting and would only
impact a branch (Postmaster be made aware) if data was retrieved from the audit store. Further if
upon retrieval of branch data from the audit store a transaction had been removed, the ‘data density’
check would highlight a missing transaction. If upon retrieval of branch data from the audit store a
transaction had been amended, the digital signature check would highlight an issue with the integrity
of the data.
ii) Branch Database privileged Oracle user operations are audited by Oracle to the SYS.AUDS$ table.
This table is extracted into audit files every night by a batch job into a directory from which the audit
archiving system extracts the data. The audit data is currently stored for 10 years. This table can be
extracted from the Audit Store by Fujitsu.
© 2016 Delc
POL00031502
POL00031502
Any amendment / deletion of data in the audit store would be visible to Fujitsu only when data is
retrieved. Upon retrieval of branch data from the audit store a transaction had been removed, the
‘data density’ check would highlight a missing transaction. If upon retrieval of branch data from the
audit store a transaction had been amended, the digital signature check would highlight an issue with
the integrity of the data.
As per the exception noted on page 3, there is a small theoretical risk of a user ‘spoofing’ the digital
signature, arising from a failure in SOD controls relating to the digital signature.
d) Whether (a) has ever actually happened?
Audit logs of super-user access in the BRDB exist. Fujitsu have confirmed where amendment /
deletion of live database tables would be identifiable from this log.
Our work has not included obtaining logs for the relevant time period and performing analytics over
them to identify any instances where this has happened, and investigate if so. Such procedures
should be theoretically possible however.
2. The key points we need to understand are whether (i) Balancing Transactions and (ii) changes by Super-
users can effect branch accounts from the perspective of the postmaster, in particular:
a) Are these changes visible to the postmaster? >.
There is no system setting which wouid flag to the Postmaster when a change had been made by a
super user.
The Transaction Log report gives the Postmaster a way of identifying Balancing Transactions, as
transactions that have been inserted can be identified as the associated user would be dispiayed as
“SUPPORTTOOLUSERQ9Q’ (i.e. not a member of staff at the Branch)
b) Can these generate a shortfall in the branch accounts?
if used in a certain way, BTs or a super-user change could theoretically cause a shortfall in branch
accounts.
c) How would this impact on the making of daily cash declarations?
Daily cash declarations are a real time report generated by a branch (counter) which queries the
BRDB live database; therefore any balancing transaction inserted into the BRDB or change of
transactional BRDB data by a super user, would automatically impact the daily cash rec report
(impact dependent on nature of BT / change).
d) I How would this impact on "monthly" branch trading balances?
The monthly Branch Trading Statement, which a Postmaster must sign off on in order to roll into the
next accounting period would also be impacted by a change of this nature.
The monthly branch trading statement, reports on data live from the BRDB, and aggregated data
from the BRDB, therefore any balancing transaction inserted into the BRDB or change of
transactional BRDB data by a super user, would automatically impact the daily cash rec report
(impact dependent on nature of BT / change).
Specific questions on the Interim Report
3. Diagram on Page 8:
a) Transfer of data from BAL to BRDB - Does this happen daily? If so when during the day? Is it overnight?
BAL is a compilation of servers used for the transfer of data from Counter to BRDB, this processing is
done in a near real time manner. As such transfer of data from BAL to BRDB is instantaneous once a
basket is complete.
i) Given the daily polling of data from which source does the Counter pull data when the
postmaster conducts an end of day cash declaration? (The above suggests the data must
be pulled from BAL as all other sources would not be up to date in real time?)
© 2016 Delc
DRAFT 74
POL00031502
POL00031502
BRDB. A request from counter is raised (via the BAL) to BRDB using pre-defined SQL
scripts at the BRDB layer to generate this cash declaration report/process. When a cash
deciaration is raised by a branch a message transfer is sent via the BAL which
communicates with the BRDB to query the live transaction tables using a pre-defined SQL
script
b) Transaction corrections generated by POL: Where does a Transaction Correction fit on this diagram?
Transaction Corrections are inserted directly into BRDB by a defined data transfer process.
c) The diagram suggests that data is held in the Audit Server for 5 days but para (iii)(b) on page 14
suggests that data is held in the BRDB for 5 days? Are both statements correct or is one a typo?
Most data is held in BRDB for approximately 5 days, (depending on specific type of data). Certain
values are also aggregated and the aggregated data held for up to 60 days to allow for real time
reports, and the monthly branch trading statement, ran by the counter to include this data if required.
Most data is held on the Audit Server for approximately 5 days, (depending on specific type of data).
4. Page 10:
a) Point F —- says POL finance staff can "input / amend" a transaction — We know they can input a
transaction but can they "amend" a transaction? If so, how?
This refers to a Transaction Correction (TC). A TC could, depending on the detail of the TC, have the
effect of ‘amending’ an existing transaction. A TC must be accepted at the counter before impacting
branch accounting.
5. Page 19:
a) What is meant by the phrase "predominantly limited to HNG-X due to previous Audit Store retention
limitations"?
Wording removed to avoid ambiguity.
b) What is meant by the phrase: "Any writes by Fujitsu Support to BRDB must be audited"?
Branch Database privileged Oracle user operations (Fujitsu Support) are audited by Oracle to the
SYS.AUDS table.
c) Atpoint “iv” — what is the difference between “Correcting” and “updating”? We did not think FJ could
“correct”, only “insert? [This point also comes up at Page 13, 1°* column of table].
ABT could, depending on the detaii of the BT, have the effect of ‘amending’ an existing transaction.
ABT can only insert, and not update or delete existing records. The possibility of a superuser
amending existing transactions does exist as highlighted above in question 1
6. BTs in relation to the SU issue:
a) Please can you explain the situation with using Balancing Transactions to solve the SU problem?
The usage of the BT tool for this purpose is not a ‘true’ BT as no data (transactions) is/are injected
into the database. However the same tool which allows a BT to be posted, is used to perform this
procedure.
The procedure is performed to update the transaction recovery table of a Stock Unit (SU) in the rare
instance when the recovery flag for a transaction gets into an inconsistent state, and needs to be
manuaily updated, to show that the transaction has been recovered by the branch.
This procedure is managed by an MSC (change request) process prior to the updates taking place.
b) Other than the one use of a BT to solve a bug, are you sure that all other uses of BTs relate to the
SU issue?
For the period data was available for and therefore reviewed (12/03/2010 — 28/05/2016).
All other uses of the tool in this period updated the specific table
BRDB_RX_RECOVERY_TRANSACTIONS' (SU issue) and did not contain INSERT statements.
c) Will the branch be aware of the SU issue?
POL00031502
POL00031502
The Branch would not be notified of the tool being used for this purpose, however this process is
generally initiated by the branch when the branch is struggling to perform this task manually using
the counter.
d) Can the SU issue ever cause a discrepancy in the branch accounts?
The usage of the tool to update the transaction recovery table of an SU does not insert / remove /
amend transactions.
7. BT audit files:
a) What do the "audit files" in relation to BTs track and show?
All usages of the too! used for inserting BTs. The logs show the actual SQL commands used to insert
the BT, and contain all fieids updated and their respective values (quantities and product ids). There
are also user timestamps which identify the user who inserted the BT.
b) How far back do the audit files go?
The audit files commence at 12/03/2010
8. FJ access to conduct a BT
a) How many staff at FU have permission to inject a BT?
31 (of these 31, 26 also have direct access to the live BRDB database and therefore could
theoretically make changes to transaction tables as described in (10b) below.
b) What is the process followed by FJ for using a BT?
The process followed by FJ is:
An error is recognised by the branch and they raise a request/call to SSC.
A TFS/Peak Incident service desk tool is then used to record incidents raised by Post masters (TFS
has subsequently been retired and incidents all 1st and 2nd line branch incidents are now recorded
in Peak Incident Management).
This issue will then be investigated by SSC. If a BT is required then this is passed to Fujitsu for
further work and solution management.
if a BT is required this is recorded on the Peak Incident ticket.
Approvals are then sought by senior members of POL before this is executed which is captured
id the ticket fequest.
c) What operational controls re there around the use of BTs at FJ?
A branch would initiate the process described in (b} above for a BT to be executed.
Senior approvals are required by POL before this process can be completed.
Use of BT tool is audited and any transactions inserted would be recognised by branch through
transactional log reports.
The BT tool is restricted to a limited number of Fujitsu personnel who are independent to the Peak
incident process.
d) What is the process followed at POL for implanting / authorising a BT (if this is out of scope, please
say and we will pick up direct with POL)?
Out of scope. Agreed POL will answer.
9. BT visibility
a) Would a BT shows in the branch accounts from a postmaster's perspective?
i) What report would a postmaster need to run?
© 2016 Delc
POL00031502
POL00031502
A Postmaster is not notified if a Balancing Transaction is inserted into the live transaction
tables.
There are various real time reports a Postmaster can run which would be affected by
something of this nature (notably the Transaction Log report, which is able to display
transactions that have been posted over the last 60 days.). Transactions in this report
would be identifiable by the user code “SUPPORTTOOLUSER99" (i.e. not a member of
staff at the Branch).
Further any Balancing Transaction impacting a branch's transactional! data would impact
declarations made by Postmasters (encouraged to do so on a daily basis) and also
declarations are required to be done in order to rollover into the next accounting period
(typically 4-5 weeks). The monthly Branch Trading Statement which a Postmaster must
sign off on in order to roll into the next accounting period would aiso be impacted by a
change of this nature which would capture summarised totals of transactional data, which
could be reconciled by branch back to the granular transaction log reports. All of the
mentioned reports are mechanisms by which the Postmaster would be made aware of a
Balancing Transaction. The reporting functionality of counters was described by Fujitsu and
this understanding was corroborated by review of technical documentation, no
walkthroughs were performed of this process.
ii) How would it be identifiable from other transactions?
Transactions in the Transaction Log report would be identifiable by the user code
“SUPPORTTOOLUSERQ9’ (i.e. not a member of staff at the Branch).
b) Cana BT by back-dated (i.e. injected into the branch accounts at an historic date)?
Whether the Balancing Transaction would be successful or not is not known by Fujitsu as it has
never been attempted. 1
POL and Deloitte are awaiting Fujitsu to provide an estimated cost / time for this walkthrough to be
performed (Cost and time. required made up primarily from creating a Suitably isolated test
environment in order to perform the walkthrough in).
Fujitsu have stated. the answer has to be yes in the sense that if the fix involves inserting a record with
an associated date then the date would be chosen as part of the design to fix the problem. The choice
of date would have to be made carefully as transactions will only be harvested from the Branch
Database for processing by back-end systems if it meets the correct selection criteria — hence the
need to test any proposed fix. . The issue is simply that we would have to invent a scenario from
‘scratch and then check that out. [ don’t see that such an exercise would add value given that we have
already carried out a walkthrough of the tool.’
c) Were BTs (or something similar) possible in Old Horizon? [See attached note from Fu]
Fujitsu have advised they have attempted to make contact to retired staff on the matter but are
unable to provide a definitive answer on processes in place pre HNG-X relating to Balancing
Transactions, only that the transaction correction too! used to inject BTs that has been used since
HNG-X implementation in 2010, was not used.
i) What controls were there around these?
Due to the response on the prevous question from Fujitsu we cannot comment on these
controls.
ji) Were they logged?
Due to the response on the prevous question from Fujitsu we cannot comment on these
controls.
© 2016 Deluitie LLP Private
POL00031502
POL00031502
10. Super-users.
a) Can Super-users only access the BRDB or can they access other servers (ie. audit server, audit
store)?
Super-users could theoretically access data at any other point in the flow of data from Counter —
Audit Store. This flow of data has been mapped by Deloitte and access rights at each point tested.
i) In Deloitte's Board Briefing Paper dated 4 June 2014, on page 2, it notes: "It is possible for
Fujitsu staff with suitably authorised privileged access to delete data from the Audit Store."
Has this issues been addressed / will it be addressed?
Yes, once data is in the audit store it cannot be amended / deleted for 7 years, as
described in (1a) above.
ii) Would deleting data from the audit store have any effect on branch accounting?
No, unless data was retrieved from the audit store which would only happen in the case of
a query being raised / investigation. It would only impact usage of this historical data for any
purposes when subsequently extracted from the audit store.
All postmaster reporting functionality is generated from the live BRDB transactional tables
(and tables which aggregate this data and store it for up to 60 days). Any amendment /
deletion of data in the audit store therefore has no effect on branch accounting and would
only impact a branch if data was retrieved from the audit store. Further if upon retrieval of
branch data from the audit store a transaction had been removed, the ‘data density’ check
would highlight a missing transaction. If upon retrieval of branch data from the audit store a
transaction had been amended, the digital signature check would highlight an issue with the
integrity of the data. As per the exception noted on page 3, there is.a small theoretical risk
of a user ‘spoofing’ the digital signature, arising froma failure in SOD controis relating to
the digital signature.
b) If a Super-user edits data in the BRDB, how w might this affect the branch accounts from the
perspective of the postmaster?
i) Where does the edited data flow to?
The edited data would remain in the BRDB transactional tables assuming that it was
entered in the correct logic
The data in this table would then follow the normal si data flow processes (i.e. BRDB > audit
server > audit store, BRDB > POLSAP, BRDB > Counter reporting etc.) if this transaction
had not already been picked up by the mechanisms which transfer transactional tables
downstream (eg. Audit track gatherer which runs every 15 minutes.)
ii) Could the edited data cause a loss in a branch's accounts?
Yes, from a branch reporting perspective any change to data in the BRDB would affect the
real time reports ran on the counter, which are used for branch accounting, specifically the
monthly Branch Trading Statement which a Postmaster must sign off on in order to roll into
the next accounting period
However if a branches data was retrieved from the audit store, any amendment to
transactional data would cause the ‘digital signature’ integrity check to fail, and Fujitsu
would be notified of this failure upon retrieval of the audit data. As per the exception noted
on page 3, there is a small theoretical risk of a user ‘spoofing’ the digital signature, arising
from a failure in SOD controls relating to the digital signature.
iii) Will the edited data be visible to the postmaster?
A Postmaster is not specifically notified if a change had been made by a ‘super-user’.
Any changes to transactional data would impact declarations made by Postmasters
(encouraged to do so on a daily basis) and also declarations are required to be done in
order to rollover into the next accounting period (typically 4-5 weeks). The monthly Branch
Trading Statement which a Postmaster must sign off on in order to roll into the next
accounting period would also be impacted by a change of this nature which would capture
summarised totals of transactional data, which could be reconciled by branch back to the
granular transaction log reports. All of the mentioned reports are mechanisms by which the
Postmaster would be made aware of any such changes.
iv) Would the edited data be visible to POL / FJ?
© 2016 Delc
POL00031502
POL00031502
Yes, as the data amendments would impact transactional records in the BRDB, and
subsequently this data would flow through to the audit store. POL / FJ would be abie to
identify this through review of audit logs as described in 1C above.
© 2016 Delsitie LLP Privaie and Confidential - Subject to Lagal Privilege - DRAFT 79
POL00031502
POL00031502
Statement of Responsibility
We take responsibility for this report which is prepared on the basis of the limitations set out below.
The matters raised in this report are only those which came to our attention during the course of our work and are
not necessarily a comprehensive statement of all the weaknesses that exist or all improvements that might be made.
Recommendations for improvements should be assessed by you for their full impact before they are implemented.
Deloitte LLP
London
October 2016
Other than as stated below, this document is confidential and prepared solely for your information and that of other
beneficiaries of our advice listed in our engagement letter. Therefore you should not, refer to or use our name or this
document for any other purpose, disclose them or refer to them in any prospectus or other document, or make them
available or communicate them to any other party. If this document contains details of an arrangement that could
result in a tax or National Insurance saving, no such conditions of confidentiality apply to the details of that
arrangement (for example, for the purpose of discussion with tax authorities). In any event, no other party is entitled
to rely on our document for any purpose whatsoever and thus we accept no liability to any other party who is shown
or gains access to this document.
© 2016 Deloitte LLP. All rights reserved.
Deloitte LLP is a limited liability partnership registered in England and Wales with registered number OC303675
and its registered office at 2 New Street Square, London EC4A 3BZ, United Kingdom.
Deloitte LLP is the United Kingdom member firm of Deloitte Touche Tohmatsu Limited (“DTTL”), a UK private
company limited by guarantee, whose member firms are legally separate and independent entities. Please see
www.deloitte.co.uk/about for a detailed description of the legal structure of DTTL and its member firms.
Member of Deloitte Touche Tohmatsu Limited