Mimicking the Familiar: Dynamic Command Generation for Information Theft Attacks in LLM Tool-Learning System
This addresses privacy risks in LLM tool-learning systems by enhancing attack adaptability, though it is incremental as it builds on existing attack methods.
The paper tackles information theft attacks in LLM tool-learning systems by proposing AutoCMD, a dynamic command generation approach that mimics familiar patterns to adapt to queries and toolchains, resulting in a +13.2% improvement in attack success rate over baselines.
Information theft attacks pose a significant risk to Large Language Model (LLM) tool-learning systems. Adversaries can inject malicious commands through compromised tools, manipulating LLMs to send sensitive information to these tools, which leads to potential privacy breaches. However, existing attack approaches are black-box oriented and rely on static commands that cannot adapt flexibly to the changes in user queries and the invocation chain of tools. It makes malicious commands more likely to be detected by LLM and leads to attack failure. In this paper, we propose AutoCMD, a dynamic attack comment generation approach for information theft attacks in LLM tool-learning systems. Inspired by the concept of mimicking the familiar, AutoCMD is capable of inferring the information utilized by upstream tools in the toolchain through learning on open-source systems and reinforcement with target system examples, thereby generating more targeted commands for information theft. The evaluation results show that AutoCMD outperforms the baselines with +13.2% $ASR_{Theft}$, and can be generalized to new tool-learning systems to expose their information leakage risks. We also design four defense methods to effectively protect tool-learning systems from the attack.