网安资讯详情 - SecLens 情报雷达

网安资讯,一网打尽。汇集权威漏洞通告与行业要闻,结合分组浏览、智能过滤、RSS订阅 和 Webhook 推送,多通道拓展您的安全情报视野。

Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-world Web Agents

来源: arxiv_cs_cr · 发布时间 2026-06-11 22:12 (UTC+08:00) · 抓取时间 2026-06-12 19:10 (UTC+08:00)

原文链接

摘要

Web agents driven by large language models (LLMs) are increasingly deployed in real-world environments, where they operate over untrusted web content and execute actions with direct consequences. This makes them vulnerable to prompt-injection attacks, in which seemingly benign content embeds adversarial instructions that manipulate agent behaviour. Existing security benchmarks adopt an \textit{attack-centric} perspective, focusing on the technical feasibility of injections while overlooking the nuanced distribution of resulting harms. In practice, however, prompt-injection risk is victim-dependent: a single exploit can produce asymmetric consequences for different stakeholders, and the same attack pattern may exhibit substantially different effectiveness depending on whom it targets. To capture these properties, we introduce \textbf{\sysname}, a \textit{stakeholder-centric} benchmark to systematically categorize and attribute harm in real-world web agent systems. It distinguishes between affected entities (e.g., user, seller, platform), decomposes the attacks into concrete objectives, and evaluates each case with complementary outcome- and process-level metrics. Our results reveal substantial and heterogeneous vulnerabilities: not a single attack objective is reliably resisted by current agents, and failures distribute across qualitatively distinct modes ranging from \emph{stealthy parasitism} (attack succeeds without disrupting the user's delegated task) to \emph{misaligned disruption} (task disrupted without attack success) and \emph{compounded failure} (both adversarial objective and task integrity simultaneously violated). These patterns are missed by conventional evaluation, highlighting the need for stakeholder-aware assessment of LLM-based agents in real-world deployments. Benchmark is available at https://github.com/StakeBench/SBC.

正文

Web agents driven by large language models (LLMs) are increasingly deployed in real-world environments, where they operate over untrusted web content and execute actions with direct consequences. This makes them vulnerable to prompt-injection attacks, in which seemingly benign content embeds adversarial instructions that manipulate agent behaviour. Existing security benchmarks adopt an \textit{attack-centric} perspective, focusing on the technical feasibility of injections while overlooking the nuanced distribution of resulting harms. In practice, however, prompt-injection risk is victim-dependent: a single exploit can produce asymmetric consequences for different stakeholders, and the same attack pattern may exhibit substantially different effectiveness depending on whom it targets. To capture these properties, we introduce \textbf{\sysname}, a \textit{stakeholder-centric} benchmark to systematically categorize and attribute harm in real-world web agent systems. It distinguishes between affected entities (e.g., user, seller, platform), decomposes the attacks into concrete objectives, and evaluates each case with complementary outcome- and process-level metrics. Our results reveal substantial and heterogeneous vulnerabilities: not a single attack objective is reliably resisted by current agents, and failures distribute across qualitatively distinct modes ranging from \emph{stealthy parasitism} (attack succeeds without disrupting the user's delegated task) to \emph{misaligned disruption} (task disrupted without attack success) and \emph{compounded failure} (both adversarial objective and task integrity simultaneously violated). These patterns are missed by conventional evaluation, highlighting the need for stakeholder-aware assessment of LLM-based agents in real-world deployments. Benchmark is available at https://github.com/StakeBench/SBC. Authors: Zihao Wang, Yiming Li, Yutong Wu, Zheyu Liu, Kangjie Chen, Fok Kar Wai, Pin-Yu Chen, Vrizlynn L. L. Thing, Bo Li, Dacheng Tao, Tianwei Zhang Categories: cs.CR, cs.AI, cs.CY, cs.HC, cs.MM PDF: https://arxiv.org/pdf/2606.13385v1 Comment: 32 pages

标签

扩展字段

{
  "arxiv_id": "2606.13385v1",
  "authors": [
    "Zihao Wang",
    "Yiming Li",
    "Yutong Wu",
    "Zheyu Liu",
    "Kangjie Chen",
    "Fok Kar Wai",
    "Pin-Yu Chen",
    "Vrizlynn L. L. Thing",
    "Bo Li",
    "Dacheng Tao",
    "Tianwei Zhang"
  ],
  "categories": [
    "cs.CR",
    "cs.AI",
    "cs.CY",
    "cs.HC",
    "cs.MM"
  ],
  "comment": "32 pages",
  "doi": null,
  "entry_id": "https://arxiv.org/abs/2606.13385v1",
  "pdf_url": "https://arxiv.org/pdf/2606.13385v1",
  "primary_category": "cs.CR",
  "search_query": "cat:cs.CR",
  "updated_at": "2026-06-11T14:12:43+00:00"
}