Cyber-Capable AI Agents: Vulnerabilities, Evaluation Containment, and Defensive Response

  • 类型:arxiv
  • 标识:2607.25379
  • 链接:http://arxiv.org/abs/2607.25379v1
  • 主分类:agent
  • 形态:method
  • 被引:0
  • 被引来源:Semantic Scholar + OpenAlex
  • S2被引:0
  • OpenAlex被引:0
  • 影响力被引:0
  • TLDR:This review synthesizes five vulnerability classes at that boundary: multi-step offensive chains, objectives that conflict with sandbox boundaries, supply-chain and credential exposure, persistent command-and-control, and the speed of automated action.
  • OpenAlex ID:W7171617418
  • OpenAlex DOI:10.48550/arxiv.2607.25379
  • DOI:10.48550/arxiv.2607.25379
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://doi.org/10.48550/arxiv.2607.25379
  • OpenAlex更新:2026-08-25
  • 副分类:evaluation
  • 待LLM分类:否
  • 标题中文:具备网络攻击能力的 AI Agent:漏洞、评估遏制与防御响应
  • TLDR中文:本文综述了该边界上的五类漏洞:多步攻击链、与沙箱边界冲突的目标、供应链与凭据暴露、持续性的 command-and-control,以及自动化行动的速度。
  • 来源文件
  • /inbox/tom/_candidates/2026-07-29-agent-rag-longcontext-candidates.json
  • /inbox/tom/_candidates/2026-07-30-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]