Reachability-Based Capability Confinement for LLM Agents under Indirect Prompt Injection

• URL: http://arxiv.org/abs/2608.30041v1
• Date: Sun, 30 Aug 2026 20:57:47 GMT
• Title: Reachability-Based Capability Confinement for LLM Agents under Indirect Prompt Injection
• Abstract summary: We present SkillGuard, a harness-level enforcement layer that treats contamination as contamination and restricts future capabilities to disconnect the resulting state from deployer-defined forbidden states.<n>We evaluate SkillGuard on four AgentDojo suites with two backend LLMs, Gemini 2.5 Flash and Llama3.3-70B, against an LLM-only No Defense baseline and three defenses at different system layers.
• Score: 8.339789704552706
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Large language model agents place outputs from external skills into their execution context, allowing attacker-controlled data to influence later privileged actions. Existing defenses mainly classify untrusted content or authorize proposed operations. They do not directly address how an agent's future authority should change once untrusted data enters its state. We present SkillGuard, a harness-level enforcement layer that treats this event as contamination and restricts future capabilities to disconnect the resulting state from deployer-defined forbidden states. Given sound skill summaries and policies, SkillGuard represents security-relevant transitions with a Skill Impact Graph, specifies admissible control over skill parameters via steerability signatures, and mediates invocations with an inline reference monitor. Following contamination, it computes weighted capability restrictions using binary, fractional, or fractional-flow strategies without auxiliary language-model inference. We evaluate SkillGuard on four AgentDojo suites with two backend LLMs, Gemini 2.5 Flash and Llama3.3-70B, against an LLM-only No Defense baseline and three defenses at different system layers: Spotlighting, CaMeL, and AttriGuard. We construct a compositional attack benchmark in which each attack combines observations individually insufficient to induce target violation and evaluate the same baselines on it. Under AgentDojo's Tool Knowledge attacks, SkillGuard eliminates attack success on three of four suites for both backends and reduces it to 4.8% and 14.3% on Slack. Against compositional attacks, it outperforms every baseline on Llama and matches the strongest baseline on Gemini at higher benign utility. Fractional-flow restriction preserves substantially more capabilities than binary restriction at the same attack success rate. Across both settings, SkillGuard adds no model calls or token overhead.

Related papers

• Privacy-Preserving Topology-Guided Safety for LLM-Based Multi-Agent Systems via Federated Graph Learning [67.04448659688579]
  FGLGuard cuts AgentDojo's ground-success rate by 43% at near-unguarded utility, zero API cost, and negligible capability loss.<n>Federation is not optional: off-the-shelf transfer collapses under distribution shift.<n>Live FGLGuard cuts AgentDojo's ground-success rate by 43% at near-unguarded utility, zero API cost, and negligible capability loss.
  arXiv  Detail & Related papers  (2026-09-02T07:57:12Z)
• SIR: Self-improving Red-teaming for Compute Use Agents [71.71987044117371]
  Computer use agents (CUAs) are vision-language models that perceive a screen and act on a real operating system through mouse, keyboard, and terminal.<n>Because they can be exposed to untrusted content while operating, they are vulnerable to indirect prompt injection (IPI)<n>We present SIR, a black box IPI attack that composes stealthy injections from a small library of reusable principles stated in plain language.
  arXiv  Detail & Related papers  (2026-08-31T03:39:43Z)
• SkillShield: Prompt-Space Security Skills for LLM Coding Agents [20.042094810707166]
  SkillShield is a system-prompt defense that synthesizes security skills offline from known attacks or recorded agent failures.<n>It protects the system by defining the security policies the model should follow during execution.<n>Across six large language models on RedCode, the default all-classes skill reduces malware-generation severity from 3.37 to 0.58 and achieves a 43.6% execution attack success rate.
  arXiv  Detail & Related papers  (2026-08-26T14:00:11Z)
• StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing [80.39644823192113]
  We propose StepGuard, a step-level guard model that can audit completed agent trajectories and check tool actions before they are executed.<n>StepGuard achieves the highest average accuracy among open-weight guard models, with performance comparable to GPT-5.4.<n>When used to guard agents on AgentDojo and AgentDyn, StepGuard reduces mean attack success rate by 77.3%.
  arXiv  Detail & Related papers  (2026-08-25T16:17:27Z)
• ElasticBack: Stealthy Conditional Backdoor in LLM-Agent Skills via Coupled Trigger-Rule Optimization [19.596331768742925]
  ElasticBack is a conditional single-skill backdoor that plants a rule R in the skill document and a benign-looking trigger T in the user query.<n>It then freezes R and evolves T against it with a stealth-constrained genetic search, so that effectiveness and stealth are optimized.<n>Experiments show that ElasticBack attains a high attack success rate at a near-zero false-positive rate with preserved clean accuracy, transfers across models, and evades deployment-time defenses.
  arXiv  Detail & Related papers  (2026-08-10T13:12:35Z)
• FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks [11.27557330957057]
  We introduce FlipGuard, a proactive defense framework that selectively perturbs model weights prior to quantization.<n>By breaking the adversary's precise alignment between weight patterns and quantization boundaries, FlipGuard suppresses backdoor activation without requiring access to training data or trigger samples.
  arXiv  Detail & Related papers  (2026-06-27T15:02:42Z)
• SecureClaw: Clawing Back Control of LLM Agents [5.952015302242554]
  Tool-using large language model (LLM) agents face two distinct security failures.<n>We present SecureClaw, a dual-boundary architecture that places authorization at the effect sink and plaintext confinement at the read boundary.<n>Across AgentDojo, AgentLeak, and Agent Security Bench (ASB), SecureClaw is the only defense we evaluate in a common harness.
  arXiv  Detail & Related papers  (2026-06-08T14:29:01Z)
• AttriGuard: Defeating Indirect Prompt Injection in LLM Agents via Causal Attribution of Tool Invocations [38.49666480491258]
  LLM agents are highly vulnerable to Indirect Prompt Injection (IPI)<n>We propose a new paradigm, action-level causal attribution, which secures agents by asking why a particular tool call is produced.<n>We instantiate this paradigm with AttriGuard, a runtime defense based on parallel counterfactual tests.
  arXiv  Detail & Related papers  (2026-03-11T13:23:46Z)
• Visual Backdoor Attacks on MLLM Embodied Decision Making via Contrastive Trigger Learning [89.1856483797116]
  We introduce BEAT, the first framework to inject visual backdoors into MLLM-based embodied agents.<n>Unlike textual triggers, object triggers exhibit wide variation across viewpoints and lighting, making them difficult to implant reliably.<n>BEAT achieves attack success rates up to 80%, while maintaining strong benign task performance.
  arXiv  Detail & Related papers  (2025-10-31T16:50:49Z)
• Malice in Agentland: Down the Rabbit Hole of Backdoors in the AI Supply Chain [82.98626829232899]
  Fine-tuning AI agents on data from their own interactions introduces a critical security vulnerability within the AI supply chain.<n>We show that adversaries can easily poison the data collection pipeline to embed hard-to-detect backdoors.
  arXiv  Detail & Related papers  (2025-10-03T12:47:21Z)
• AdvEvo-MARL: Shaping Internalized Safety through Adversarial Co-Evolution in Multi-Agent Reinforcement Learning [78.5751183537704]
  AdvEvo-MARL is a co-evolutionary multi-agent reinforcement learning framework that internalizes safety into task agents.<n>Rather than relying on external guards, AdvEvo-MARL jointly optimize attackers and defenders.
  arXiv  Detail & Related papers  (2025-10-02T02:06:30Z)

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.