AgentSafe: Safeguarding Large Language Model-based Multi-agent Systems via Hierarchical Data Management

2025-03-06

AI Summary

To address security threats—including data leakage and unauthorized access—in LLM-driven multi-agent systems (MAS), this paper proposes HieroSec, a hierarchical data management framework comprising two core components: ThreatSieve (a threat-filtering module) and HierarCache (a hierarchical caching mechanism). HieroSec introduces the first systematic defense mechanism tailored for agent memory, integrating information-level authorization, authenticated inter-agent communication, dynamic access control, and adversarially robust memory management, while supporting interoperability across multiple LLM backends. Experimental results demonstrate >80% adversarial defense success rate across diverse large language models, scalability to hundreds of agents, and favorable trade-offs among security assurance, computational overhead, and runtime stability. To our knowledge, HieroSec is the first fine-grained, scalable memory security solution specifically designed for MAS.

Technology Category

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

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Security and Privacy: Large-scale security measurements

Abstract

Large Language Model based multi-agent systems are revolutionizing autonomous communication and collaboration, yet they remain vulnerable to security threats like unauthorized access and data breaches. To address this, we introduce AgentSafe, a novel framework that enhances MAS security through hierarchical information management and memory protection. AgentSafe classifies information by security levels, restricting sensitive data access to authorized agents. AgentSafe incorporates two components: ThreatSieve, which secures communication by verifying information authority and preventing impersonation, and HierarCache, an adaptive memory management system that defends against unauthorized access and malicious poisoning, representing the first systematic defense for agent memory. Experiments across various LLMs show that AgentSafe significantly boosts system resilience, achieving defense success rates above 80% under adversarial conditions. Additionally, AgentSafe demonstrates scalability, maintaining robust performance as agent numbers and information complexity grow. Results underscore effectiveness of AgentSafe in securing MAS and its potential for real-world application.

Problem

• Enhances security in multi-agent systems using hierarchical data management.
• Protects against unauthorized access and data breaches in LLM-based systems.
• Introduces systematic defense mechanisms for agent memory and communication.

Innovation

• Hierarchical data management enhances MAS security.
• ThreatSieve verifies information authority, prevents impersonation.
• HierarCache defends against unauthorized access, memory poisoning.