arXiv cs.MA· Diego Gosmar, Deborah A. Dahl·· 11 小时前AI 评分31
Sentinel Agents：为多智能体系统构建安全可信的智能体 AI 防护层
Sentinel Agents for Secure and Trustworthy Agentic AI in Multi-Agent Systems
AI 导读
论文提出 Sentinel Agents，一种监控 LLM 多智能体系统内部通信的专用安全层，通过规则检查、行为分析和 LLM 语义审查识别恶意或错误消息，并向 Coordinator Agent 报告以执行隔离等策略。
正文 · 原文
Abstract:Multi-agent systems (MAS) built on large language models (LLMs) increasingly share a common conversational space, in which a single adversarial or erroneous message can propagate from one agent to many. This paper proposes Sentinel Agents, a dedicated security layer that monitors inter-agent communication with rule-based checks, behavioral analysis and LLM-based semantic inspection, and reports to a Coordinator Agent that enforces policy, for example by quarantining misbehaving agents. We describe the architecture, its deployment patterns and monitoring layers, and we evaluate a continuous-listener prototype on a balanced corpus of 1,000 synthetic messages (500 attacks across five families and 500 benign messages) with two local LLMs. LLM inspection with Gemma 4 12B reached F1 0.86 at a false-positive rate of 0.11, against 0.38 for rules alone. Combining components by logical OR did not improve on it, while a cascade in which only high-precision rules act on their own matched it. Calibrating the thresholds of the more conservative Qwen3 8B on a development split raised its held-out F1 from 0.56 to 0.80, and on multi-turn conversations the behavioral check detected patterns the rules missed but could not tell legitimate repetition from probing. The value of combining detectors therefore depends on the model, on its calibration and on the aggregation rule, and the results yield practical guidance for deploying Sentinel Agents.
| Comments: | 23 pages, 11 figures. v2: the preliminary evaluation of v1 is replaced by a new evaluation on a balanced 1,000-message corpus with ablation, local LLM benchmark, aggregation and calibration analyses; paper restructured; references updated |
| Subjects: | Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2509.14956 [cs.AI] |
| (or arXiv:2509.14956v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2509.14956 arXiv-issued DOI via DataCite |
Submission history
From: Diego Gosmar [view email]
[v1]
Thu, 18 Sep 2025 13:39:59 UTC (2,000 KB)
[v2]
Thu, 5 Nov 2026 17:05:56 UTC (2,018 KB)
来源：arXiv cs.MA · arxiv.org