analiza aiqbench
This research addresses the critical challenge of defending LLM-based multi-agent systems against evolving adversarial attacks in open-world environments. By implementing a co-evolutionary framework that adapts to distribution shifts and behavioral drift, OpenEvoShield demonstrates robust detection capabilities without suffering from catastrophic forgetting. The study provides essential insights for securing autonomous agent deployments where static defenses are insufficient against dynamic threat landscapes.
OpenEvoShield: Dual Non-Stationary Continual Defense for Open-World Multi-Agent System Attacks
arXiv:2607.19351v1 Announce Type: new Abstract: LLM-based multi-agent systems (LLM-MAS) are increasingly deployed in safety-critical applications, where adversaries inject malicious instructions through inter-agent communication to propagate harmful behaviors. Unlike static threats, these attacks are doubly dynamic: adversaries refine injection strategies against deployed defenses while normal-agent behavior drifts with system expansion. Existing defenses treat deployment as a closed-world problem and degrade rapidly once either distribution shifts beyond training coverage. We propose OpenEvoShield, a co-evolutionary continual defense framework for LLM-MAS. An asymmetric rate controller (M1) decouples fast attack-side and slow normal-side learning rates from dual drift signals. A normal-boundary updater (M2) maintains a dynamic behavioral boundary at the slow rate, while an EWC-regularized policy ensemble (M3) fast-adapts without catastrophic forgetting. An energy-based multi-granularity detector (M4) fuses node-, subgraph-, and graph-level evidence to classify novel attacks as out-of-distribution. Experiments over 100 deployment rounds across five benchmarks and four MAS topologies show that OpenEvoShield outperforms static and continual baselines, detecting most previously unseen attacks while keeping false positive rates low.
Abstract:LLM-based multi-agent systems (LLM-MAS) are increasingly deployed in safety-critical applications, where adversaries inject malicious instructions through inter-agent communication to propagate harmful behaviors. Unlike static threats, these attacks are doubly dynamic: adversaries refine injection strategies against deployed defenses while normal-agent behavior drifts with system expansion. Existing defenses treat deployment as a closed-world problem and degrade rapidly once either distribution shifts beyond training coverage. We propose OpenEvoShield, a co-evolutionary continual defense framework for LLM-MAS. An asymmetric rate controller (M1) decouples fast attack-side and slow normal-side learning rates from dual drift signals. A normal-boundary updater (M2) maintains a dynamic behavioral boundary at the slow rate, while an EWC-regularized policy ensemble (M3) fast-adapts without catastrophic forgetting. An energy-based multi-granularity detector (M4) fuses node-, subgraph-, and graph-level evidence to classify novel attacks as out-of-distribution. Experiments over 100 deployment rounds across five benchmarks and four MAS topologies show that OpenEvoShield outperforms static and continual baselines, detecting most previously unseen attacks while keeping false positive rates low.
Artificial Intelligence (cs.AI)
arXiv:2607.19351 [cs.AI]
(or arXiv:2607.19351v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2607.19351 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 252 cuvinte
