ThreatFabric Explains Dual-Model Behavioural Analytics for Detecting Banking Fraud

Summary

ThreatFabric describes combining per-user behaviour models with fraud-trained models to detect scams, APP fraud and account takeover in banking sessions, reporting detection-rate improvements of 20% or more when combined with transaction signals.

Key points

• An Identity Model learns each customer's interaction patterns and flags deviations that may indicate account or device takeover, remote access, or scam manipulation.
• A Fraudster Model is trained on confirmed fraud and scam cases and looks for patterns such as excessive copy-and-paste, unusual navigation, bot-like speed, hesitation, and irregular data entry.
• ThreatFabric says combining both models can reduce false positives from normal behavioural changes and improve fraud detection rates by 20% or more when fused with transaction signals.
• The approach is intended to detect social engineering, impersonation, investment and romance scams, including APP fraud involving a genuine but manipulated customer.
• ThreatFabric says its system minimizes PII collection, does not use physical biometrics, and analyses keystroke timing and rhythm without recording typed content.

Vendors

Products

Related Articles

• Threat Intelligence Roundup Covers Pentagon Breach, FBI Operations and Hirudo Trojandarkowl.comThreatFabric
• Weekly Security Roundup: Supermarket Attacks, Ransomware and Major Patchesdatabreachtoday.comThreatFabric