AI-BASED BEHAVIOURAL ANOMALY DETECTION FOR CYBERSECURITY IN DIGITAL BANKING

Abstract

India's banking system has moved rapidly from branch-based services toward mobile banking, internet banking, instant payments and data-intensive financial services. This change has created a large security surface in which fraud, phishing, malware, credential theft, account takeover, mule accounts and attacks on digital payment channels can occur at high speed. Artificial intelligence (AI), especially machine learning, can help banks examine large volumes of transactions, identify unusual behaviour, rank alerts, detect suspicious networks and support faster cybersecurity response. This research paper examines the role of AI in cybersecurity in India's banking sector using a documentary and secondary-data approach. The study uses Reserve Bank of India (RBI) evidence on reported bank frauds for 2022-23 to 2024-25, RBI material on artificial intelligence and supervisory analytics, The Indian Computer Emergency Response Team -CERT-In cyber incident requirements, and established academic literature on anomaly detection, explainable AI, adversarial machine learning and cyber threat detection. The RBI data show that reported fraud cases rose sharply in 2023-24 and remained substantial in 2024-25, while the distribution of cases and amounts differed across bank groups. In 2024-25, private sector banks reported 14,233 fraud cases and public sector banks reported 6,935; however, the public sector group accounted for the largest reported amount among the bank groups shown in the RBI table. The paper explains why AI should be treated as a security capability rather than as a replacement for human investigation. It proposes a practical architecture combining data quality, behavioural analytics, graph analysis, threat intelligence, explainability, human review, model monitoring, privacy protection and incident response. The study concludes that responsible AI can strengthen banking cybersecurity when it is governed through clear accountability, independent validation, secure data practices and continuous testing.