AI-INSPIRED BEHAVIOURAL INTRUSION DETECTION WITH DECEPTIVE ACCESS CONTROL FOR WEB APPLICATIONS

Web applications are increasingly exposed to unauthorized access attempts, administrative probing, and credential misuse attacks.Existing intrusiondetection systemsprimarily rely on static rules and signature-based techniques,which are ineffective againstadaptive attackers that imitate legitimate user behaviour.This paper presents an AI-inspiredbehavioural intrusion detection system with deceptive access control for web applications.Theproposed system continuously monitors user behaviour in real time and evaluates accesspatterns using predefined behavioural indicators, including request frequency anomalies, role-based access violations, and repeated access to restricted resources.A dynamic risk-scoringmechanism is employed to detect suspicious activity without requiring machine learning modeltraining, enabling lightweight and real-time deployment.Instead of immediately blocking suspicious users, deceptive access control is applied by redirecting unauthorized users to decoyadministrative interfaces, protecting critical resources while allowing attacker behaviourobservation. Repeated malicious actions beyond a definedthreshold result in automaticblocking.The system is implemented using Flask andMySQL, and experimental resultsdemonstrate effective detection withreduced false positives in real-world webenvironments.