Cowpox: Towards the Immunity of VLM-based Multi-Agent Systems

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wu, Yutong, Zhang, Jie, Li, Yiming, Zhang, Chao, Guo, Qing, Lukas, Nils, Zhang, Tianwei
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911103390318592
author Wu, Yutong
Zhang, Jie
Li, Yiming
Zhang, Chao
Guo, Qing
Lukas, Nils
Zhang, Tianwei
author_facet Wu, Yutong
Zhang, Jie
Li, Yiming
Zhang, Chao
Guo, Qing
Lukas, Nils
Zhang, Tianwei
contents Vision Language Model (VLM)-based agents are stateful, autonomous entities capable of perceiving and interacting with their environments through vision and language. Multi-agent systems comprise specialized agents who collaborate to solve a (complex) task. A core security property is robustness, stating that the system should maintain its integrity under adversarial attacks. However, the design of existing multi-agent systems lacks the robustness consideration, as a successful exploit against one agent can spread and infect other agents to undermine the entire system's assurance. To address this, we propose a new defense approach, Cowpox, to provably enhance the robustness of multi-agent systems. It incorporates a distributed mechanism, which improves the recovery rate of agents by limiting the expected number of infections to other agents. The core idea is to generate and distribute a special cure sample that immunizes an agent against the attack before exposure and helps recover the already infected agents. We demonstrate the effectiveness of Cowpox empirically and provide theoretical robustness guarantees.
format Preprint
id arxiv_https___arxiv_org_abs_2508_09230
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Cowpox: Towards the Immunity of VLM-based Multi-Agent Systems
Wu, Yutong
Zhang, Jie
Li, Yiming
Zhang, Chao
Guo, Qing
Lukas, Nils
Zhang, Tianwei
Multiagent Systems
Artificial Intelligence
Vision Language Model (VLM)-based agents are stateful, autonomous entities capable of perceiving and interacting with their environments through vision and language. Multi-agent systems comprise specialized agents who collaborate to solve a (complex) task. A core security property is robustness, stating that the system should maintain its integrity under adversarial attacks. However, the design of existing multi-agent systems lacks the robustness consideration, as a successful exploit against one agent can spread and infect other agents to undermine the entire system's assurance. To address this, we propose a new defense approach, Cowpox, to provably enhance the robustness of multi-agent systems. It incorporates a distributed mechanism, which improves the recovery rate of agents by limiting the expected number of infections to other agents. The core idea is to generate and distribute a special cure sample that immunizes an agent against the attack before exposure and helps recover the already infected agents. We demonstrate the effectiveness of Cowpox empirically and provide theoretical robustness guarantees.
title Cowpox: Towards the Immunity of VLM-based Multi-Agent Systems
topic Multiagent Systems
Artificial Intelligence
url https://arxiv.org/abs/2508.09230