ALERT: A Closed-Loop SOC Agent for Web and API Security with CVE-Based Intelligence and Adaptive Anomaly Detection
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Web and Application Programming Interface (API) services are central to modern digital infrastructure but remain vulnerable to injection attacks, authentication bypasses, and zero-day exploits. Traditional Security Operations Center (SOC) tools rely on static, signature-based pipelines that generate high false positives, struggle with concept drift, and provide limited decision support. This paper presents ALERT (Agentic Learning for Event-driven Response and Threat Detection), a lightweight, GPU-free (CPU-only), closed-loop SOC agent with a human-in-the-loop (HITL) that integrates Common Vulnerabilities and Exposures (CVE) severity prediction, log anomaly detection, and Language Model (LLM)-assisted semantics into a cost-sensitive decision framework. ALERT ingests multimodal inputs, including vulnerability feeds, runtime logs, and analyst annotations, and dynamically adjusts responses (ALLOW, MONITOR, or BLOCK) through online policy adaptation. Evaluation on a synthetic dataset of 10,000 Apache-style logs achieved a precision of 0.991, a recall of 0.983, and an F1-score of 0.987, detecting 100 percent of critical threats while keeping false positives below 0.5 percent. Designed for scalability and transparency, ALERT’s modular architecture generalizes to Cyber-Physical Systems (CPS) and Internet of Things (IoT) telemetry, supporting practical, adaptive, and explainable SOC automation.
Fields of Study & Subjects
Network Security and Intrusion Detection
Web Application Security Vulnerabilities
Software System Performance and Reliability
Publication details
- Journal / source
- International Conference on Consumer Electronics
- Pages
- 1–6
- Publication date
- 3 February 2026
- Language
- EN