Sentinel: A Trust-Aware API Security Model Using Behavioural Analysis and Human-in-the-Loop AI Retraining

This paper presents Sentinel, a trust-aware API security middleware that enhances traditional security models by incorporating continuous behavioral analysis. Unlike conventional systems that rely on static authentication methods, Sentinel dynamically evaluates trust using a hybrid approach that combines rule-based detection, machine-learning-based anomaly detection, and human-in-the-loop feedback. The system is implemented using FastAPI and ASGI middleware architecture, ensuring real-time request analysis with minimal latency. Experimental results demonstrate significant improvements in detecting advanced threats, including credential stuffing, geo-jump attacks, and low-rate DDoS attacks. Sentinel provides a scalable, efficient solution for modern API security challenges that aligns with zero-trust architecture principles.