arXiv:2609.06140 · Agent 智能体
Counter-Swarm Doctrine: Containing Coordinated Agent Intrusions
反集群 doctrine:遏制协同化 Agent 入侵
Counter-Swarm Doctrine: Containing Coordinated Agent Intrusions
- 类型:arxiv
- 标识:2609.06140
- 链接:https://arxiv.org/abs/2609.06140
- 主分类:agent
- 形态:method
- TLDR:Agents can turn shared infrastructure into a channel for coordinated intrusion. The Hugging Face incident and a separate public-wiki investigation show why a security assessment may need evidence from several executions and the artifacts they leave behind. We argue that the operational unit of defence should be a revisable coordination episode linking observed transfers, task authority, and response history. The central research problem is prospective episode discovery: finding which actions belong together before an evaluator supplies their membership. We define unsanctioned coordination rela
- 待LLM分类:否
- 标题中文:反集群 doctrine:遏制协同化 Agent 入侵
- TLDR中文:Agent 可将共享基础设施变为协同入侵的通道。Hugging Face 事件与一项公开 wiki 调查表明,安全评估为何可能需要来自多次执行及其遗留产物的证据。我们主张防御的作战单元应是一个可修订的协同 episode,将观察到的传输、任务授权与响应历史相链接。核心研究问题是前瞻性 episode 发现:在评估者给出成员归属之前,找出哪些动作彼此关联。我们定义未经授权的协同关系……
- 来源文件:
- /inbox/tom/_candidates/2026-09-10-agent-rag-longcontext-candidates.json