A Survey on Agentic Security: Applications, Threats and Defenses

2025-10-07

AI Summary

Autonomous LLM-based agents present dual-use capabilities in cybersecurity—offering novel offensive and defensive applications while simultaneously introducing unprecedented security risks. Method: We propose the first tripartite framework for agent security—encompassing application scenarios, threat modeling, and defense mechanisms—systematically synthesizing over 150 scholarly works. Employing bibliometric analysis, cross-domain integrative assessment, and taxonomy-driven modeling, we construct a comprehensive, multimodal agent security classification system spanning model-level and multimodal operational contexts. Contribution/Results: Our framework identifies critical vulnerabilities—including architectural dependencies, goal hijacking, and multi-step reasoning flaws—alongside corresponding defense gaps. We deliver a structured knowledge graph that clarifies prevailing research trends, persistent challenges, and critical voids. This work establishes a foundational theoretical basis, a standardized evaluation paradigm, and a principled roadmap for technological advancement in LLM agent security.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Multiagent Systems: Modeling other Agents
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Agentic search

Abstract

The rapid shift from passive LLMs to autonomous LLM-agents marks a new paradigm in cybersecurity. While these agents can act as powerful tools for both offensive and defensive operations, the very agentic context introduces a new class of inherent security risks. In this work we present the first holistic survey of the agentic security landscape, structuring the field around three interdependent pillars: Applications, Threats, and Defenses. We provide a comprehensive taxonomy of over 150 papers, explaining how agents are used, the vulnerabilities they possess, and the countermeasures designed to protect them. A detailed cross-cutting analysis shows emerging trends in agent architecture while revealing critical research gaps in model and modality coverage.

Problem

• Surveying autonomous LLM-agent cybersecurity applications and risks
• Analyzing inherent security threats in agentic AI systems
• Developing defense taxonomies for agent vulnerabilities and countermeasures

Innovation

• First holistic survey of agentic security landscape
• Comprehensive taxonomy of over 150 research papers
• Cross-cutting analysis reveals critical research gaps