Mainstream poisoning attacks on large language models (LLMs) typically set a fixed trigger in the input instance and specific responses for triggered queries. However, the fixed trigger setting (e.g., unusual words) may be easily detected by human detection, limiting the effectiveness and practicality in real-world scenarios. To enhance the stealthiness of the trigger, we present a poisoning attack against LLMs that is triggered by a generation/output condition-token limitation, which is a commonly adopted strategy by users for reducing costs. The poisoned model performs normally for output without token limitation, while becomes harmful for output with limited tokens. To achieve this objective, we introduce BrieFool, an efficient attack framework. It leverages the characteristics of generation limitation by efficient instruction sampling and poisoning data generation, thereby influencing the behavior of LLMs under target conditions. Our experiments demonstrate that BrieFool is effective across safety domains and knowledge domains. For instance, with only 20 generated poisoning examples against GPT-3.5-turbo, BrieFool achieves a 100% Attack Success Rate (ASR) and a 9.28/10 average Harmfulness Score (HS) under token limitation conditions while maintaining the benign performance.
The increasing demand for customized Large Language Models (LLMs) has led to the development of solutions like GPTs. These solutions facilitate tailored LLM creation via natural language prompts without coding. However, the trustworthiness of third-party custom versions of LLMs remains an essential concern. In this paper, we propose the first instruction backdoor attacks against applications integrated with untrusted customized LLMs (e.g., GPTs). Specifically, these attacks embed the backdoor into the custom version of LLMs by designing prompts with backdoor instructions, outputting the attacker's desired result when inputs contain the pre-defined triggers. Our attack includes 3 levels of attacks: word-level, syntax-level, and semantic-level, which adopt different types of triggers with progressive stealthiness. We stress that our attacks do not require fine-tuning or any modification to the backend LLMs, adhering strictly to GPTs development guidelines. We conduct extensive experiments on 4 prominent LLMs and 5 benchmark text classification datasets. The results show that our instruction backdoor attacks achieve the desired attack performance without compromising utility. Additionally, we propose an instruction-ignoring defense mechanism and demonstrate its partial effectiveness in mitigating such attacks. Our findings highlight the vulnerability and the potential risks of LLM customization such as GPTs.