G-Safeguard: A Topology-Guided Security Lens and Treatment on LLM-based Multi-agent Systems

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Shilong, Zhang, Guibin, Yu, Miao, Wan, Guancheng, Meng, Fanci, Guo, Chongye, Wang, Kun, Wang, Yang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912233576988672
author Wang, Shilong
Zhang, Guibin
Yu, Miao
Wan, Guancheng
Meng, Fanci
Guo, Chongye
Wang, Kun
Wang, Yang
author_facet Wang, Shilong
Zhang, Guibin
Yu, Miao
Wan, Guancheng
Meng, Fanci
Guo, Chongye
Wang, Kun
Wang, Yang
contents Large Language Model (LLM)-based Multi-agent Systems (MAS) have demonstrated remarkable capabilities in various complex tasks, ranging from collaborative problem-solving to autonomous decision-making. However, as these systems become increasingly integrated into critical applications, their vulnerability to adversarial attacks, misinformation propagation, and unintended behaviors have raised significant concerns. To address this challenge, we introduce G-Safeguard, a topology-guided security lens and treatment for robust LLM-MAS, which leverages graph neural networks to detect anomalies on the multi-agent utterance graph and employ topological intervention for attack remediation. Extensive experiments demonstrate that G-Safeguard: (I) exhibits significant effectiveness under various attack strategies, recovering over 40% of the performance for prompt injection; (II) is highly adaptable to diverse LLM backbones and large-scale MAS; (III) can seamlessly combine with mainstream MAS with security guarantees. The code is available at https://github.com/wslong20/G-safeguard.
format Preprint
id arxiv_https___arxiv_org_abs_2502_11127
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle G-Safeguard: A Topology-Guided Security Lens and Treatment on LLM-based Multi-agent Systems
Wang, Shilong
Zhang, Guibin
Yu, Miao
Wan, Guancheng
Meng, Fanci
Guo, Chongye
Wang, Kun
Wang, Yang
Cryptography and Security
Machine Learning
Multiagent Systems
Large Language Model (LLM)-based Multi-agent Systems (MAS) have demonstrated remarkable capabilities in various complex tasks, ranging from collaborative problem-solving to autonomous decision-making. However, as these systems become increasingly integrated into critical applications, their vulnerability to adversarial attacks, misinformation propagation, and unintended behaviors have raised significant concerns. To address this challenge, we introduce G-Safeguard, a topology-guided security lens and treatment for robust LLM-MAS, which leverages graph neural networks to detect anomalies on the multi-agent utterance graph and employ topological intervention for attack remediation. Extensive experiments demonstrate that G-Safeguard: (I) exhibits significant effectiveness under various attack strategies, recovering over 40% of the performance for prompt injection; (II) is highly adaptable to diverse LLM backbones and large-scale MAS; (III) can seamlessly combine with mainstream MAS with security guarantees. The code is available at https://github.com/wslong20/G-safeguard.
title G-Safeguard: A Topology-Guided Security Lens and Treatment on LLM-based Multi-agent Systems
topic Cryptography and Security
Machine Learning
Multiagent Systems
url https://arxiv.org/abs/2502.11127