Abstract

Artificial Intelligence agents have emerged as autonomous systems capable of addressing cybersecurity's scalability challenges through automated threat detection, response, and security assessment. However, the landscape of AI agent applications in cybersecurity remains fragmented across domains and publication venues. This systematic literature review analyzes 57 peer-reviewed studies to provide comprehensive mapping of AI agent research in cybersecurity. Through thematic analysis, we identify and categorize agent applications into two primary domains: defensive agents (86% of papers) encompassing intrusion detection, malware analysis, security management, and threat intelligence; and offensive agents (16% of papers) including penetration testing, fuzzing, and attack simulation. Our analysis reveals that intrusion detection dominates with 70% of papers, while offensive applications remain underexplored. We contribute a comprehensive taxonomy of AI agent roles, systematic analysis of technical approaches within each category, and identification of underexplored research areas to guide future investigations.

Download

Share

COinS