Analysis And Detection Of Network Security Failures In Banking Systems Using Machine Learning

Abstract

Introduction: Digital banking depends on interconnected networks that support internet and mobile banking, automated teller machines, payment gateways, application programming interfaces, and core banking services. This connectivity improves access and transaction speed, but it also creates opportunities for credential abuse, malware communication, denial-of-service activity, unauthorized access, and data exhilaration. This paper reformulates the supplied case study into an IEEE-style technical article and presents a machine-learning (ML) framework for identifying suspicious network behaviour in banking environments. Publicly reported figures reproduced in the source document provide context: the Reserve Bank of India (RBI) reported 23,953 bank-fraud cases involving ₹36,014 crore in 2024–2025 for frauds of ₹1 lakh and above, while card/internet fraud represented 56.5% of reported cases by area of operation [1]. The proposed workflow combines traffic and authentication telemetry, pre-processing, feature extraction, supervised classification, anomaly detection, risk scoring, and analyst-led response. Model-performance figures in the source material are illustrative rather than experimentally validated; accordingly, no empirical superiority claim is made. The paper also discusses evaluation measures, false alarms, operational limitations, and steps required for a reproducible implementation.