Red Teaming Language Model Detectors with Language Models
This work addresses the safety and ethical risks of LLM misuse by highlighting vulnerabilities in detection systems, which is incremental as it builds on prior detection methods.
The paper tackles the problem of detecting LLM-generated text by testing the robustness of existing detectors under adversarial attacks, showing that attacks using synonym replacement and prompt alteration effectively compromise all detectors with plausible generations.
The prevalence and strong capability of large language models (LLMs) present significant safety and ethical risks if exploited by malicious users. To prevent the potentially deceptive usage of LLMs, recent works have proposed algorithms to detect LLM-generated text and protect LLMs. In this paper, we investigate the robustness and reliability of these LLM detectors under adversarial attacks. We study two types of attack strategies: 1) replacing certain words in an LLM's output with their synonyms given the context; 2) automatically searching for an instructional prompt to alter the writing style of the generation. In both strategies, we leverage an auxiliary LLM to generate the word replacements or the instructional prompt. Different from previous works, we consider a challenging setting where the auxiliary LLM can also be protected by a detector. Experiments reveal that our attacks effectively compromise the performance of all detectors in the study with plausible generations, underscoring the urgent need to improve the robustness of LLM-generated text detection systems.