Industry
Banking
Banking Services Transformation: Unlocking Exponential Value
Highlights
- Financial institutions are facing increasingly sophisticated, AI-driven scams that bypass traditional perimeter security controls, where a single fraudulent event can quickly escalate from account takeover to illicit money laundering.
- The solution lies in building a real-time behavioural biometrics platform that captures distinct user interaction patterns across digital channels to construct a highly personalised and unique biometric signature.
- By comparing live session activities against this established customer baseline, the system dynamically calculates risk scores to detect subtle behavioural anomalies without disrupting legitimate user journeys.
- Based on the calculated threat level, the decisioning engine triggers automated actions—either allowing low-risk actions, stepping up authentication, or blocking high-risk transactions.
- Behavioural biometrics provides a critical layer of defence against complex scams, establishing itself as a foundational element of future financial security and digital trust.
On this page
On this pageinpage
How can banks use digital DNA to stop fraud in real-time?
In the banking and financial services industry, fraud continues to grow in both scale and complexity, resulting in significant losses for financial institutions and their customers. Globally, loss due to scams and frauds registered a compounded annual growth of 9.2% to touch $579.4 billion in 2025 compared to 2023, according to Nasdaq Verafin's 2026 Global Financial Crime Report.1
The rising incidence of fraud is resulting in legal action against banks, leading to financial and reputational loss. For example, in June 2026 , Australia's Federal Court ordered a global bank to pay a penalty of A$35 million2 after it admitted to failures in protecting customers from scams. Financial crime can no longer be classified into separate traditional categories such as cyber, fraud, or anti-money laundering (AML). This is because they rarely occur as separate events—criminals do not operate within such boundaries. A stolen credential may first appear as a suspicious login, then quickly become an account takeover, and within minutes, the funds can be routed through multiple mule accounts or high-value ecommerce purchases all within minutes, before conventional fraud detection mechanisms can act.
We believe that banks and financial institutions must build a digital behavioural biometrics platform underpinned by a shared data and analytics layer to detect and prevent fraud and money-laundering events in real-time. While this is a daunting task, digital banking channels already capture the signals needed to detect fraud, often identifying fraud risk before the core transaction monitoring system does. These signals from the digital banking channel can be leveraged to build a unique digital DNA or behavioural biometric profiles or signatures for individual customers. When customers initiate transactions, their digital behaviour can be compared to the established profile, proactively identifying deviations or anomalies to prevent fraudulent transactions from going through.
What challenges arise in building a digital behavioural biometrics platform?
Building a digital behavioural biometrics platform will come with its own challenges; however, effectively addressing them is crucial to success.
Operational siloes: The primary challenge is establishing a common approach through which customer signals can be interpreted consistently across fraud, cyber, and AML incidents. Additionally, siloed operations and a lack of interoperability between different systems prevent banks and financial institutions from accurately analysing and interpreting signals.
Behavioural data gaps: Most banks already collect basic information such as event timestamps, user identifiers, navigation paths, authentication events, and interaction histories required to build a digital behavioural biometrics platform. However, they lack the ability to capture the more distinctive behavioural signals, such as mouse movement patterns, keystroke cadence, gesture characteristics, swipe dynamics, and interaction velocity, which may demand additional capabilities.
Governance and privacy: Behavioural biometrics can be perceived as intrusive if not governed appropriately. Data minimisation principles should be incorporated into the design from the outset. Where persistent customer identifiers are required, tokenisation or pseudonymisation techniques should be used to reduce privacy risk while preserving analytical value.
How is a digital behavioural biometric signature created?
A digital biometric profile is a statistical portrait of how a specific customer accesses and interacts with digital channels. Individual biometric signals in isolation are unlikely to unlock value. Digital indicators such as telemetry, device intelligence, authentication behaviour, transaction activity, and customer context must combine to form a coherent view of behavioural risk and provide early warning of cyber compromise, fraud, or money-laundering. Grouping the signals into digital indicator categories helps arrive at the digital behavioural biometric signature while ensuring the design is easily understood by business, technology, and compliance teams because each category represents a visible aspect of customer behaviour (see Figure 1).
Figure 1: Digital indicator categories converge to form a composite digital behavioural biometric signature for each customer
Let us examine the steps involved in building a digital behavioural biometric signature for individual customers.
Baseline digital behaviour
Banks and financial institutions must observe customers’ behaviour during a 60-90-day enrolment period. The baseline must be adequately flexible to adapt to genuine changes in customer behaviour such as relocation, a changed work schedule, or a new phone, and so on. But a sudden cluster of changes should be treated as a high-risk event and trigger scrutiny. This distinction between gradual drift and abrupt multi-signal deviation is what makes the approach practical.
Adopt multiple modelling techniques
A single model will be insufficient to detect digital behavioural deviations. A combination of models will be needed to detect various risks: anomaly detection models for unknown attacks, supervised learning models for known fraud patterns, graph analytics for mule networks, and drift detection models for genuine behavioural change. Financial institutions must design an ensemble model that combines the outputs of each model to arrive at a composite risk score that gives a comprehensive view of behavioural risk.
Score risk to detect anomalous digital behaviour in real-time
Proactive detection involves a continuous comparison between the live session and the established digital behavioural biometric signature or profile, which is done by the decision engine. A single odd event may be explainable, but several deviations occurring together will raise the risk score and trigger a stronger response from the platform (see Figure 2).
Figure 2: Real-time anomaly-to-action decision flow for login, navigation, and transaction events
Digital behavioural deviations typically indicate anomalies that fall into three categories. The digital behavioural biometrics platform must then trigger an automated response to stop the transaction and prevent fraud (see Table 1).
Table 1: Anomaly classification and indicative response
Based on its similarity to a customer’s established digital behavioural profile, the score estimates the likelihood of the session having been initiated by the enrolled customer. The output can be expressed on a 0–100 scale and mapped to graded actions with thresholds finetuned for each customer (see Table 2). Reason codes should be assigned to decisions to help in understanding why the session was challenged, held, or allowed.
Table 2: Illustrative composite risk score interpretation
What are the building blocks of a fraud and anti-money laundering (FRAML) architecture?
Individual financial institutions may build the digital biometrics behavioural platform differently depending on organisation-specific requirements. However, a digital behavioural biometric architecture must incorporate certain key components to support foundational capabilities (see Figure 3).
- Analytics engine to evaluate digital indicators of customers’ online transactions or interactions.
- Streaming event bus to support real-time ingestion of digital data.
- Schema registry to enforce contracts and tokenise personally identifiable information (PII) on entry.
- Repository to store each customer’s baseline biometric signature. The low-latency online biometric features power live scoring, while offline biometric features feed model training and periodic recalibration.
- Decision engine along with applicable rules to operationalise the biometric signature.
- Policy layer to translate the composite score into a concrete decision, such as allow, step-up, hold, block, or open a FRAML case when thresholds are breached.
- Feedback mechanism to assess whether an alert was a false positive and facilitate a continuous flow of insights back into the system to improve precision through retraining of models, recalibration of thresholds, and refreshing of the customer's baseline.
Figure 3: Digital behavioural biometrics architecture
Why is behavioural biometrics the future of digital trust?
The importance of digital behavioural biometric capabilities is set to increase as AI continues to lower the barriers for fraudsters to execute sophisticated frauds at scale. Traditional controls are inadequate to protect against AI-generated phishing campaigns, voice cloning, synthetic identities, and highly personalised social-engineering attacks. As attackers become more effective at acquiring valid credentials and successfully bypassing authentication checkpoints, customers’ unique digital DNA or digital biometric behaviour will emerge as a critical differentiator in verifying their antecedents. The primary benefit is in timing; the bank can take action before the funds leave a customer’s account. Digital behavioural biometric signatures may ultimately become as important to digital trust as passwords were to the first generation of online banking.
Banks must start by parallelly running proofs of concept (PoCs) on high-risk use cases, baseline performance, and compare the AI recommendation with actual outcomes. This will help measure the benefits without disrupting business-as-usual. Banks that act decisively and build a digital behavioural biometrics platform will gain a competitive advantage, as behavioural models improve with every interaction observed, creating a history that cannot be built retrospectively.
References
1 Nasdaq Verafin's 2026 Global Financial Crime Report, March 2026, November 2026, https://verafin.com/wp-content/uploads/2026/03/global-financial-crime-report-2026-nasdaq-verafin-20260316.pdfv
2 ASIC, Newsroom, Federal Court orders $35 million penalty against HSBC for scam protection failures, June 2026, November 2026, https://www.asic.gov.au/about-asic/news-centre/find-a-media-release/2026-releases/26-127mr-federal-court-orders-35-million-penalty-against-hsbc-for-scam-protection-failures