LITMUS: Benchmarking Behavioral Jailbreaks of LLM Agents in Real OS Environments

• URL: http://arxiv.org/abs/2605.10779v1
• Date: Mon, 11 May 2026 16:14:04 GMT
• Title: LITMUS: Benchmarking Behavioral Jailbreaks of LLM Agents in Real OS Environments
• Abstract summary: behavior jailbreak is where an adversary induces an agent to execute dangerous OS-level operations with irreversible consequences.<n>Existing benchmarks either evaluate safety at the semantic layer alone, missing physical-layer harms, or fail to isolate test cases.<n>We present LITMUS, a benchmark addressing both gaps via a semantic-physical dual verification mechanism and OS-level state rollback.
• Score: 26.85543164204341
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The rapid proliferation of LLM-based autonomous agents in real operating system environments introduces a new category of safety risk beyond content safety: behavior jailbreak, where an adversary induces an agent to execute dangerous OS-level operations with irreversible consequences. Existing benchmarks either evaluate safety at the semantic layer alone, missing physical-layer harms, or fail to isolate test cases, letting earlier runs contaminate later ones. We present LITMUS (LLM-agents In-OS Testing for Measuring Unsafe Subversion), a benchmark addressing both gaps via a semantic-physical dual verification mechanism and OS-level state rollback. LITMUS comprises 819 high-risk test cases organized into one harmful seed subset and six attack-extended subsets covering three adversarial paradigms (jailbreak speaking, skill injection, and entity wrapping), plus a fully automated multi-agent evaluation framework judging behavior at both conversational and OS-level physical layers. Evaluation across frontier agents reveals three findings: (1) current agents lack effective safety awareness, with strong models (e.g., Claude Sonnet 4.6) still executing 40.64% of high-risk operations; (2) agents exhibit pervasive Execution Hallucination (EH), verbally refusing a request while the dangerous operation has already completed at the system level, invisible to every prior semantic-only framework; and (3) skill injection and entity wrapping attacks achieve high success rates, exposing pronounced agent vulnerabilities. LITMUS provides the first standardized platform for reproducible, physically grounded behavioral safety evaluation of LLM agents in real OS environments.

Related papers

• MobileWorldSafety: Benchmarking GUI Agent Safety Against Environmental Injection Attacks in Android Apps [26.50307679295416]
  We introduce MobileWorldSafety, a benchmark of risk tasks built on real Android applications.<n>For each task, we define a verifiable risk indicator over the final system state.<n>This distinguishes safety failures from capability failures and enables objective and reproducible assessment.
  arXiv  Detail & Related papers  (2026-08-18T11:33:22Z)
• AgentLens: Interpretable Safety Steering via Mechanistic Subspaces for Multi-Turn Coding Agent [53.82005364479556]
  Coding agents based on large language models (LLMs) demonstrate remarkable autonomous capabilities.<n>Existing safety mechanisms mainly rely on external guardrails.<n>We propose AgentLens, a white-box defense framework that performs runtime safety detection and representation-level mitigation.
  arXiv  Detail & Related papers  (2026-06-21T21:13:55Z)
• SeClaw: Spec-Driven Security Task Synthesis for Evaluating Autonomous Agents [87.26967184869198]
  SeClaw is a framework that combines specification-driven security task synthesis with execution-based security evaluation for Autonomous agents.<n> benchmark covers risks arising from resources, user tasks, environments, and intrinsic agent behaviors.
  arXiv  Detail & Related papers  (2026-06-01T14:23:42Z)
• When the Manual Lies: A Realistic Benchmark to Evaluate MCP Poisoning Attacks for LLM Agents [43.702131498259384]
  This paper systematically investigates Tool Description Poisoning (TDP), a novel semantic attack.<n>In TDP, malicious instructions are not embedded in a tool's executable code, but rather covertly injected into its descriptive metadata.<n>This work provides the first specialized security benchmark tailored for TDP, offering essential insights for securing the cognitive and planning layers of advanced agentic systems.
  arXiv  Detail & Related papers  (2026-05-22T08:34:48Z)
• The Dark Side of LLMs: Agent-based Attacks for Complete Computer Takeover [0.0]
  Large Language Model (LLM) agents and multi-agent systems introduce security vulnerabilities that extend beyond traditional content generation to system-level compromises.<n>This paper presents a comprehensive evaluation of the LLMs security used as reasoning engines within autonomous agents.<n>We show how different attack surfaces and trust boundaries can be leveraged to orchestrate such takeovers.
  arXiv  Detail & Related papers  (2025-07-09T13:54:58Z)
• OpenAgentSafety: A Comprehensive Framework for Evaluating Real-World AI Agent Safety [58.201189860217724]
  We introduce OpenAgentSafety, a comprehensive framework for evaluating agent behavior across eight critical risk categories.<n>Unlike prior work, our framework evaluates agents that interact with real tools, including web browsers, code execution environments, file systems, bash shells, and messaging platforms.<n>It combines rule-based analysis with LLM-as-judge assessments to detect both overt and subtle unsafe behaviors.
  arXiv  Detail & Related papers  (2025-07-08T16:18:54Z)
• IS-Bench: Evaluating Interactive Safety of VLM-Driven Embodied Agents in Daily Household Tasks [30.535665641990114]
  We present IS-Bench, the first multi-modal benchmark designed for interactive safety.<n>It features 161 challenging scenarios with 388 unique safety risks instantiated in a high-fidelity simulator.<n>It facilitates a novel process-oriented evaluation that verifies whether risk mitigation actions are performed before/after specific risk-prone steps.
  arXiv  Detail & Related papers  (2025-06-19T15:34:46Z)
• AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions [64.85086226439954]
  We present SAFE, a benchmark for assessing the safety of embodied VLM agents on hazardous instructions.<n> SAFE comprises three components: SAFE-THOR, SAFE-VERSE, and SAFE-DIAGNOSE.<n>We uncover systematic failures in translating hazard recognition into safe planning and execution.
  arXiv  Detail & Related papers  (2025-06-17T16:37:35Z)
• AgentVigil: Generic Black-Box Red-teaming for Indirect Prompt Injection against LLM Agents [54.29555239363013]
  We propose a generic black-box fuzzing framework, AgentVigil, to automatically discover and exploit indirect prompt injection vulnerabilities.<n>We evaluate AgentVigil on two public benchmarks, AgentDojo and VWA-adv, where it achieves 71% and 70% success rates against agents based on o3-mini and GPT-4o.<n>We apply our attacks in real-world environments, successfully misleading agents to navigate to arbitrary URLs, including malicious sites.
  arXiv  Detail & Related papers  (2025-05-09T07:40:17Z)
• SafeAgentBench: A Benchmark for Safe Task Planning of Embodied LLM Agents [58.65256663334316]
  We present SafeAgentBench -- the first benchmark for safety-aware task planning of embodied LLM agents in interactive simulation environments.<n>SafeAgentBench includes: (1) an executable, diverse, and high-quality dataset of 750 tasks, rigorously curated to cover 10 potential hazards and 3 task types; (2) SafeAgentEnv, a universal embodied environment with a low-level controller, supporting multi-agent execution with 17 high-level actions for 9 state-of-the-art baselines; and (3) reliable evaluation methods from both execution and semantic perspectives.
  arXiv  Detail & Related papers  (2024-12-17T18:55:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.