Decision Models as Retrieval Gates Against Injection and Pollution
- Published
- Record updated
Summary
The authors test whether open-weight decision models can screen retrieved web passages for hidden injected instructions and synthetic misleading pages. On a 1,000-question benchmark, a Kev-4B gate reading each passage's HTML cut attack success from 27.9% with no defence to 13.6% with published mitigations and to 5.7% with the gate added, at 76 ms per passage and 4 of 186 benign pages wrongly flagged. On a 200-query pollution testbed, filtering the candidate pool before selecting the top ten kept fake pages out of answers (0.5% exposure) and held accuracy at 83%, against 28% with no filter.
Mitigation
Gate retrieved passages with a released open-weight decision model such as Kev-4B that reads each passage's HTML, applied to the candidate pool before the top ten are chosen; the published mitigations of sanitisation, normalisation and citation prompting are also tested as baselines.
Related items
- MediumCVE-2026-123851: JetBrains Junie GitHub Action prompt injection bypass via HTML encodingSimilar attack · NVD/CVE Database
- InfoMultimodal-SEA: Benchmarking AI Agent Security under Multimodal Social Engineering AttacksSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
- MediumResearchers Bypass ChatGPT Lockdown Mode to Exfiltrate Chats Without User InteractionSimilar attack · The Hacker News
- InfoTrueFoundry & Straiker | AI Gateway Security for AgentsSimilar attack · Straiker Blog
- InfoThe Agent Smith Dilemma: Governing Agentic IdentitiesSimilar attack · SecurityWeek