paper-with-me

홈 › Papers

Chain-of-Scrutiny: Detecting Backdoor Attacks for Large Language Models

2024-06-10 · Xi Li, Ruofan Mao, Yusen Zhang, Renze Lou, Chen Wu, Jiaqi Wang

Large Language Models (LLMs), especially those accessed via APIs, have demonstrated impressive capabilities across various domains. However, users without technical expertise often turn to (untrustworthy) third-party services, such as prompt engineering, to enhance their LLM experience, creating vulnerabilities to adversarial threats like backdoor attacks. Backdoor-compromised LLMs generate malicious outputs to users when inputs contain specific "triggers" set by attackers. Traditional defense strategies, originally designed for small-scale models, are impractical for API-accessible LLMs due to limited model access, high computational costs, and data requirements. To address these limitations, we propose Chain-of-Scrutiny (CoS) which leverages LLMs' unique reasoning abilities to mitigate backdoor attacks. It guides the LLM to generate reasoning steps for a given input and scrutinizes for consistency with the final output -- any inconsistencies indicating a potential attack. It is well-suited for the popular API-only LLM deployments, enabling detection at minimal cost and with little data. User-friendly and driven by natural language, it allows non-experts to perform the defense independently while maintaining transparency. We validate the effectiveness of CoS through extensive experiments on various tasks and LLMs, with results showing greater benefits for more powerful LLMs.

📄 PDF Abstract BibTeX arXiv:2406.05948

Code (1)

lixi1994/CoS 공식 구현 pytorch

Tasks

Prompt Engineering

Methods 이 논문이 사용한 방법론

SET Dynamic Sparse Training method where weight mask is updated randomly periodically

Similar Papers 제목 키워드 기반

On Evaluating Neural Network Backdoor Defenses

2020-10-23 · Akshaj Veldanda, Siddharth Garg

Deep neural networks (DNNs) demonstrate superior performance in various fields, including scrutiny and security. However, recent studies have shown that DNNs are vulnerable to backdoor attacks. Several defenses were prop…

CLIBE: Detecting Dynamic Backdoors in Transformer-based NLP Models

2024-09-02 · Rui Zeng, Xi Chen, Yuwen Pu, Xuhong Zhang 외

Backdoors can be injected into NLP models to induce misbehavior when the input text contains a specific feature, known as a trigger, which the attacker secretly selects. Unlike fixed words, phrases, or sentences used in …

Text ClassificationText Generation

BadChain: Backdoor Chain-of-Thought Prompting for Large Language Models

2024-01-20 · Zhen Xiang, Fengqing Jiang, Zidi Xiong, Bhaskar Ramasubramanian 외

Large language models (LLMs) are shown to benefit from chain-of-thought (COT) prompting, particularly when tackling tasks that require systematic reasoning processes. On the other hand, COT prompting also poses new vulne…

Backdoor Attack

Shortcuts Everywhere and Nowhere: Exploring Multi-Trigger Backdoor Attacks

2024-01-27 · Yige Li, Jiabo He, Hanxun Huang, Jun Sun 외

Backdoor attacks have become a significant threat to the pre-training and deployment of deep neural networks (DNNs). Although numerous methods for detecting and mitigating backdoor attacks have been proposed, most rely o…

Model Supply Chain Poisoning: Backdooring Pre-trained Models via Embedding Indistinguishability

2024-01-29 · Hao Wang, Shangwei Guo, Jialing He, Hangcheng Liu 외

Pre-trained models (PTMs) are widely adopted across various downstream tasks in the machine learning supply chain. Adopting untrustworthy PTMs introduces significant security risks, where adversaries can poison the model…

Backdoor Attack