Targeted Bit-Flip Attacks on LLM-Based Agents

TL;DR

This paper introduces Flip-Agent, a novel targeted bit-flip attack framework that exploits hardware faults to manipulate large language model-based agents, revealing significant security vulnerabilities in multi-stage AI systems.

Contribution

The work presents the first targeted BFA framework for LLM-based agents, demonstrating its effectiveness and exposing new security risks in complex AI pipelines.

Findings

Flip-Agent outperforms existing BFAs on real-world tasks

Significant vulnerability identified in LLM-based agent systems

Manipulates both outputs and tool invocations

Abstract

Targeted bit-flip attacks (BFAs) exploit hardware faults to manipulate model parameters, posing a significant security threat. While prior work targets single-step inference models (e.g., image classifiers), LLM-based agents with multi-stage pipelines and external tools present new attack surfaces, which remain unexplored. This work introduces Flip-Agent, the first targeted BFA framework for LLM-based agents, manipulating both final outputs and tool invocations. Our experiments show that Flip-Agent significantly outperforms existing targeted BFAs on real-world agent tasks, revealing a critical vulnerability in LLM-based agent systems.