Searcharxiv⌕ Search

arXiv subjects

Doron Shavit

Publications and source records attributed to Doron Shavit.

2 recordsLinked to original sources

Least-Privilege Auditing of Agent Tool Calls with Intent Graphs and Reasoning-Guided Escalation

LLM agents that call tools are exposed to indirect prompt injection: instructions hidden in tool outputs can steer an agent into calls the user never asked for. We present a runtime audit that derives an executable intent graph from the user's objective and the deployment's tool catalog with one LLM call per task, and checks every tool call against it with a deterministic loop that makes no LLM call (median 5.7 microseconds per call). Two further mechanisms recover utility. Lookups of the user's own state bypass the audit, because the harm of an attack lies in its write or outward step. The calls that remain flagged are escalated to a reasoning guardian, in the spirit of tool-integrated reasoning models such as START: a reasoning model that sees only the objective, the plan, the call history without tool outputs, and the flagged call, and can verify facts with a restricted Python-expression tool. A current agent showed no baseline attack success, so end-to-end validation remains open.

cs.CR↗

Recursive language models for jailbreak detection: a procedural defense for tool-augmented agents

Jailbreak prompts are a practical and evolving threat to large language models (LLMs), particularly in agentic systems that execute tools over untrusted content. Many attacks exploit long-context hiding, semantic camouflage, and lightweight obfuscations that can evade single-pass guardrails. We present RLM-JB, an end-to-end jailbreak detection framework built on Recursive Language Models (RLMs), in which a root model orchestrates a bounded analysis program that transforms the input, queries worker models over covered segments, and aggregates evidence into an auditable decision. RLM-JB treats detection as a procedure rather than a one-shot classification: it normalizes and de-obfuscates suspicious inputs, chunks text to reduce context dilution and guarantee coverage, performs parallel chunk screening, and composes cross-chunk signals to recover split-payload attacks. On AutoDAN-style adversarial inputs, RLM-JB achieves high detection effectiveness across three LLM backends (ASR/Recall 92.5-98.0%) while maintaining very high precision (98.99-100%) and low false positive rates (0.0-2.0%), highlighting a practical sensitivity-specificity trade-off as the screening backend changes.

cs.CR↗