Results show that ActGuard reduces attack success rates to a level comparable to state-of-the-art defenses while maintaining task utility close to the no-attack setting, achieving a favorable security-utility trade-off.
Abstract
Large language model (LLM) agents interact with external environments through tool invocation, but tool outputs can also expose them to indirect prompt injection (IPI) attacks. Existing defenses mainly rely on prompt hardening, content filtering, pre-generated plans, or permission constraints. These approaches often struggle with complex tasks or over-sanitize external content, making it difficult to balance security and utility. The key challenge is therefore to preserve execution flexibility while precisely identifying and removing the malicious content that actually induces unsafe actions. To address this challenge, we propose ActGuard, a pre-execution action auditing framework. Rather than judging whether external content is inherently suspicious, ActGuard assesses whether it causes the current action to deviate from a locally reasonable expectation. At each step, ActGuard predicts the tools likely to be used by the upcoming action and constructs a local tool prior without constraining the execution trajectory. Before execution, it compares the candidate action against this prior and performs tool-level contrastive analysis and parameter-level evidence localization to identify deviations in tool selection and action parameters. A verifier then examines the localized evidence, masks only spans confirmed as malicious, and regenerates the action from the sanitized context. This design preserves legitimate planning flexibility while minimizing information loss from indiscriminate filtering. We evaluate ActGuard on challenging benchmarks for tool-using agents. Results show that ActGuard reduces attack success rates to a level comparable to state-of-the-art defenses while maintaining task utility close to the no-attack setting, achieving a favorable security-utility trade-off. Our code is publicly available at: https://github.com/binzhwang/ActGuard.
APEX is an active defense that enforces both at this boundary from a single authorization contract compiled before untrusted execution, and applies uniformly across capability units without attack-specific policies or taint tracking.
Xin-Ran Zheng, Xin-Fan Guo, Zhi-Qian Hao et al.· 1 citation
This work argues that agentic risk is progressive: it can enter at four loci of the agent control loop--skill admission, invocation-time intent, execution-time effect, and post-action consequence--while a denied dangerous objective can reappear across surface forms, tools, or turns.
Kai Wang, Zeming Wei, Biaojie Zeng et al.· 0 citations
ROPE is presented, which is anchored in a structural notion of trust, and holds attack success rate to 1.6--2.6 while retaining 82--100% of undefended clean utility, significantly exceeding state-of-the-art system-level defenses in utility while attaining comparable or better security.
Xinhang Ma, Chao-Wei Xiao, William Yeoh et al.· 3 citations· ⚡2
This work forms attack and defense as partially observed state control in SEAD, deriving their design requirements from this shared execution process and constructs an environment-verifiable dataset integrating controlled initial states, replayable tool environments, and task-specific executable checks.
Xin-Jie Shen, Jun-Ran Wang, Rong-Zhe Wei et al.· 0 citations
Results show that execution-boundary authorization grounded in trusted user intent and runtime evidence can restrict unauthorized Tool Calls induced by skill injection, while maintaining high benign-task completion.
Ji-Hun Han, Ye-Jin Jang, Byung Il Kwak et al.· 0 citations
SkillGuard is presented, 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 and preserves substantially more capabilities than binary restriction at the same attack success rate.
Wu-Jie Xiong, Rabimba Karanjai, Yang Lu et al.· 2 citations