REAL TIME BEHAVIOURAL BIOMETRIC FRAUD DETECTION

Abstract

The rapid growth of digital platforms and online transactions has significantly increased the risk of fraud and unauthorized access, making traditional authentication methods insufficient for modern security needs. Conventional systems rely heavily on static credentials such as passwords and onetime passwords, which are vulnerable to phishing, credential theft, and social engineering attacks. To overcome these limitations, this work presents a real-time behavioural biometric fraud detection system that continuously monitors user interactions throughout a session. The system captures behavioural features such as keystroke dynamics, typing speed, mouse movements, and session activity, which are unique to each user. These features are processed and analyzed using machine learning techniques, particularly unsupervised learning models like Isolation Forest, to identify deviations from normal behaviour patterns. Unlike traditional systems, the proposed approach enables continuous authentication without interrupting user experience. When abnormal behaviour is detected, the system triggers alerts or additional verification mechanisms to prevent fraudulent access. The system is designed to be scalable, efficient, and adaptable to changing user behaviour over time. It can be integrated into banking, e-commerce, and enterprise applications to enhance security and reduce financial losses. Experimental results indicate that the system achieves high detection accuracy while maintaining real-time performance. Overall, the proposed solution provides a reliable and intelligent approach for fraud detection by combining behavioural biometrics with machine learning techniques.