StealthRank: LLM Ranking Manipulation via Stealthy Prompt Optimization
This work addresses vulnerabilities in LLM-driven ranking systems, which is an incremental but important domain-specific security issue.
The paper tackles the problem of adversarial ranking manipulations in LLM-driven information retrieval systems by introducing StealthRank, a method that uses energy-based optimization and Langevin dynamics to generate stealthy prompts, resulting in consistent outperformance of state-of-the-art baselines in effectiveness and stealth.
The integration of large language models (LLMs) into information retrieval systems introduces new attack surfaces, particularly for adversarial ranking manipulations. We present $\textbf{StealthRank}$, a novel adversarial attack method that manipulates LLM-driven ranking systems while maintaining textual fluency and stealth. Unlike existing methods that often introduce detectable anomalies, StealthRank employs an energy-based optimization framework combined with Langevin dynamics to generate StealthRank Prompts (SRPs)-adversarial text sequences embedded within item or document descriptions that subtly yet effectively influence LLM ranking mechanisms. We evaluate StealthRank across multiple LLMs, demonstrating its ability to covertly boost the ranking of target items while avoiding explicit manipulation traces. Our results show that StealthRank consistently outperforms state-of-the-art adversarial ranking baselines in both effectiveness and stealth, highlighting critical vulnerabilities in LLM-driven ranking systems. Our code is publicly available at $\href{https://github.com/Tangyiming205069/controllable-seo}{here}$.