Abstract
Modern applications rely heavily on APIs, which makes them a major target for cyberattacks. Traditional security models focus mainly on perimeter protection and often fail to provide continuous verification after a user is authenticated. This work presents an AI-powered Zero Trust API Security Platform designed to monitor, detect, and automatically respond to suspicious API behavior in real time. The system combines rule-based behavioral analytics with an unsupervised machine learning model (Isolation Forest) to identify both known attack patterns and previously unseen anomalies. The platform implements multi-layer security using JWT authentication, role-based access control, IP rate limiting, and token revocation mechanisms. A behavioral analytics engine evaluates each request against multiple threat detection rules covering injection attacks, credential abuse, endpoint scanning, and automated attack tools. In parallel, a Python-based ML service analyzes request features to assign anomaly scores that trigger automated blocking when risk exceeds a defined threshold. Real- time WebSocket alerts provide administrators with continuous visibility into threat activity. The system was developed using Spring Boot, React, PostgreSQL, MongoDB, and Python Flask and validated through simulated attack scenarios. Results show that the combined rule-based and ML approach enables fast detection and automated remediation of diverse API threats. During implementation, challenges related to feature selection and balancing detection sensitivity with false positives were observed, highlighting the need for adaptive security models in dynamic API environments. This work demonstrates that integrating Zero Trust principles with AI-driven behavioral monitoring can provide a practical and scalable approach to securing modern API-driven applications.