AI-Powered Identity Threat Detection and Response (ITDR): Enhancing IAM with Behavioral Biometrics and Agentic AI

This paper presents an AI-powered Identity Threat Detection and Response (ITDR) framework that integrates behavioral biometrics—unique patterns in human activities to verify identity that focus on how a person does something rather than what they are. The proposed system captures user-specific behavioral traits such as keystroke dynamics, mouse movement, and navigation patterns to detect phishing in realtime both before and after the user has logged in, not like traditional IAM systems that work only while logging in. A working prototype was developed using Python (Flask), JavaScript, and a Random Forest classifier. The system was evaluated against traditional IAM setups across five threat scenarios: credential theft, phishing, session hijacking, insider threats, and normal user behavior. Results show an improvement in detection accuracy, reduction in false positives, and enhanced session integrity through continuous authentication.