Moving the FICO Falcon fraud platform to an active/active footprint across two Australian AWS Regions, with data residency, bank-controlled keys, recovery measured in seconds and compliance evidence that is produced continuously rather than attested annually. 

The Opportunity

What the client needed to unlock

As one of Australia's leading banks, every card and payment transaction its customers make passes through the FICO Falcon fraud platform, which scores the transaction in real time against behavioural profiles and detection rules before authorisation completes. Fraud detection is therefore both a Tier-1 operational control and the custodian of some of the bank's most sensitive information: transaction histories, customer behaviour and the logic that protects them.

As the bank advanced its data centre exit strategy, it needed to unlock a fraud platform that could scale with transaction growth, change faster and run at lower cost, while meeting a regulatory bar that was rising at the same time. APRA CPS 234 required information security controls commensurate with the sensitivity of the data and demonstrable evidence that those controls operate effectively, including where assets are hosted by third parties. CPS 230, effective  July 1,  2025, required the bank to keep this critical operation within defined tolerance levels through severe disruption. The Privacy Act 1988 and the Australian Privacy Principles, in particular APP 8 on cross-border disclosure, governed where customer data could reside and who could access it, and the bank's security baseline drew on the Australian Signals Directorate (ASD)Essential Eight and the Information Security Registered Assessors Program assessment of AWS Australian Regions.

The timing was fixed by the data centre exit, and the stakes were clear. Falcon ran on an end-of-support estate of AIX scoring and case-management servers, Oracle on AIX and a proprietary profile store, with no economically viable cross-site recovery for a control that cannot be allowed to stop. For a CPS 234-regulated entity that was a control gap to be remediated and evidenced, not simply a modernisation item. Success meant a platform on which Australian residency, bank-held key custody, controlled operator access and tested recovery were properties of the architecture, delivered without disrupting the authorisation path or the upstream host and MQ-based fraud-check flows during transition.

The Strategy

Our consulting-led, AI-powered approach

Wipro shaped a sovereign-by-design migration with the bank and AWS, built around three design choices that connected directly to the regulatory and operational opportunity.

Keep everything in Australia, and make that unpassable. Rather than relying on contractual assurances, Wipro deployed FICO Falcon 7.0 across the AWS Melbourne and Sydney Regions with AWS Organizations Service Control Policies and AWS Control Tower data-residency guardrails blocking resource creation anywhere else, so primary data, recovery copies, backups and audit logs cannot leave the country even by accident. Encryption of the Oracle, Couchbase, EBS and EFS tiers uses AWS Key Management Services (KMS) customer-managed keys under bank-owned key policies, with AWS Key Management Services (KMS) External Key Store available to anchor keys in bank-held HSMs so that decryption authority can be withdrawn by the bank at any time. Compute runs on the AWS Nitro System, removing operator access to instance memory and storage; all Wipro and bank access is federated through the bank's directory with least privilege, MFA and recorded privileged sessions; and platform vendor support is performed through bank-brokered, recorded sessions with no fraud data, profiles or rules leaving the Australian Regions.

Design resilience to be measured against the business service line agreement (SLA). Because a fraud check sits inside the authorisation path, the bank's real resilience requirement was not 'recover within hours' but 'never breach the 100 ms end-to-end fraud-check SLA'. Wipro chose a fully active/active topology in which both Regions serve live scoring traffic simultaneously, with health-check-driven routing to rebalance traffic, Oracle replication and Couchbase Cross Data Center Replication keeping profiles and case data synchronised, and the shared rules repository on Amazon EFS replicated cross-Region. Within each Region, every tier spans three Availability Zones behind load balancers and Auto Scaling. The 100 ms SLA became the acceptance criterion for every resilience exercise, and timestamped routing, replication-lag and latency records from those exercises are retained as CPS 230 evidence.

Generate compliance evidence as a by-product of operating the platform. Organisation-wide AWS CloudTrail, AWS Config resource history, Virtual Private Cloud (VPC) VPC Flow Logs and database audit logs are enabled in both Regions and retained in Australian S3 with Object Lock, with controls mapped to CPS 234, the ASD Essential Eight and the IRAP-assessed control set. AWS Config conformance packs and AWS Security Hub turn residency, encryption and access controls into a continuously evaluated posture that can be exported for APRA supervisors. Powered by Wipro Intelligence™, Wipro's migration factory used automated discovery and dependency mapping of the legacy estate, including the MQ fraud-check flows and upstream host interfaces, to sequence the cutover so that co-existence with on-premises systems was maintained and the data centre exit milestone protected.

The Proof

Measurable outcomes, powered by Wipro Intelligence™

The programme delivered measurable results across sovereignty, resilience, compliance, agility and cost:
  • 100% of fraud data, profiles, rules, backups and audit logs resident in Australian AWS Regions, with Region lock enforced by SCPs and Control Tower guardrails and encryption keys under bank-owned policies.
  • Recovery time and recovery point objectives of seconds for the loss of an entire Region, through active/active operation across Melbourne and Sydney with continuous data replication, versus no viable cross-site recovery on the legacy estate.
  • 100 ms end-to-end fraud-check SLA maintained across both Regions and held as the pass/fail criterion for Availability Zone and Region resilience exercises.
  • Three Availability Zones per Region for every platform tier, so an Availability Zone failure is absorbed without invoking cross-Region rebalancing; platform designed to 99.99% availability.
  • Seven-year immutable retention of CloudTrail, Config, flow-log, database-audit and backup records in Australian S3 with Object Lock, aligned to APRA record-keeping and CPS 234 evidentiary requirements.
  • Up to 3× faster deployment cycles for detection rules and platform changes through infrastructure as code and CI/CD.
  • Estimated 60% expansion in detection and monitoring coverage across monetary and non-monetary transactions, enabled by horizontally scalable scoring and profile tiers that remove the legacy volume ceiling.
  • 30–40% infrastructure optimisation versus the on-premises estate, with early value realised inside 12 months and the data centre exit milestone met.

The Impact

What this unlocked for the business

The bank's most sensitive operational control now runs on a platform where residency, key custody and operator access are enforced by architecture rather than asserted in contracts. When APRA asks where fraud data is, who can reach it and whether the bank can keep scoring transactions through a Region loss, the answers come from live platform posture and retained exercise records rather than from point-in-time attestations, which changes the nature of the supervisory conversation from assurance to evidence.

Operationally, fraud detection has moved from being a constraint to being a capability the bank can grow. Because scoring and profile tiers scale horizontally and rule changes deploy in a fraction of the time, the fraud team can respond to new attack patterns and extend monitoring into channels and transaction types that the legacy platform could not economically cover. Because the recovery path is automatic and measured against the same 100 ms SLA the business already lives by, resilience is no longer a separate programme but a property of normal operations.

Beyond the fraud platform itself, the bank now holds a proven sovereign landing pattern for regulated, latency-sensitive workloads on AWS: Australian-only residency enforced by guardrails, bank-held keys, Nitro-based isolation, continuous compliance evidence and active/active resilience. That pattern is reusable for the next critical system to leave the data centre, which means the exit strategy can proceed on the strength of a demonstrated model rather than a new argument each time. The client, Wipro and AWS are showing that a regulated Australian bank can place a Tier-1 control on public cloud and come out with more sovereignty, more resilience and more agility than it had on premises