multi-step-trojan-attack

Tag

Cards List
#multi-step-trojan-attack

From Prompt Injection to Persistent Control: Defending Agentic Harness Against Trojan Backdoors

Hugging Face Daily Papers · 2026-05-29 Cached

This paper introduces multi-step trojan attacks against local LLM agents, where malicious prompts are embedded across multiple operations to bypass existing defenses. It proposes ClawTrojan benchmark and DASGuard defense to detect and mitigate such attacks.

0 favorites 0 likes
← Back to home

Submit Feedback