Breaking PEFT Limitations: Leveraging Weak-to-Strong Knowledge Transfer for Backdoor Attacks in LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhao, Shuai, Gan, Leilei, Guo, Zhongliang, Wu, Xiaobao, Jia, Yanhao, Xiao, Luwei, Nguyen, Cong-Duy, Tuan, Luu Anh
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909680211591168
author Zhao, Shuai
Gan, Leilei
Guo, Zhongliang
Wu, Xiaobao
Jia, Yanhao
Xiao, Luwei
Nguyen, Cong-Duy
Tuan, Luu Anh
author_facet Zhao, Shuai
Gan, Leilei
Guo, Zhongliang
Wu, Xiaobao
Jia, Yanhao
Xiao, Luwei
Nguyen, Cong-Duy
Tuan, Luu Anh
contents Despite being widely applied due to their exceptional capabilities, Large Language Models (LLMs) have been proven to be vulnerable to backdoor attacks. These attacks introduce targeted vulnerabilities into LLMs by poisoning training samples and full-parameter fine-tuning (FPFT). However, this kind of backdoor attack is limited since they require significant computational resources, especially as the size of LLMs increases. Besides, parameter-efficient fine-tuning (PEFT) offers an alternative but the restricted parameter updating may impede the alignment of triggers with target labels. In this study, we first verify that backdoor attacks with PEFT may encounter challenges in achieving feasible performance. To address these issues and improve the effectiveness of backdoor attacks with PEFT, we propose a novel backdoor attack algorithm from the weak-to-strong based on Feature Alignment-enhanced Knowledge Distillation (FAKD). Specifically, we poison small-scale language models through FPFT to serve as the teacher model. The teacher model then covertly transfers the backdoor to the large-scale student model through FAKD, which employs PEFT. Theoretical analysis reveals that FAKD has the potential to augment the effectiveness of backdoor attacks. We demonstrate the superior performance of FAKD on classification tasks across four language models, four backdoor attack algorithms, and two different architectures of teacher models. Experimental results indicate success rates close to 100% for backdoor attacks targeting PEFT.
format Preprint
id arxiv_https___arxiv_org_abs_2409_17946
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Breaking PEFT Limitations: Leveraging Weak-to-Strong Knowledge Transfer for Backdoor Attacks in LLMs
Zhao, Shuai
Gan, Leilei
Guo, Zhongliang
Wu, Xiaobao
Jia, Yanhao
Xiao, Luwei
Nguyen, Cong-Duy
Tuan, Luu Anh
Cryptography and Security
Artificial Intelligence
Computation and Language
Despite being widely applied due to their exceptional capabilities, Large Language Models (LLMs) have been proven to be vulnerable to backdoor attacks. These attacks introduce targeted vulnerabilities into LLMs by poisoning training samples and full-parameter fine-tuning (FPFT). However, this kind of backdoor attack is limited since they require significant computational resources, especially as the size of LLMs increases. Besides, parameter-efficient fine-tuning (PEFT) offers an alternative but the restricted parameter updating may impede the alignment of triggers with target labels. In this study, we first verify that backdoor attacks with PEFT may encounter challenges in achieving feasible performance. To address these issues and improve the effectiveness of backdoor attacks with PEFT, we propose a novel backdoor attack algorithm from the weak-to-strong based on Feature Alignment-enhanced Knowledge Distillation (FAKD). Specifically, we poison small-scale language models through FPFT to serve as the teacher model. The teacher model then covertly transfers the backdoor to the large-scale student model through FAKD, which employs PEFT. Theoretical analysis reveals that FAKD has the potential to augment the effectiveness of backdoor attacks. We demonstrate the superior performance of FAKD on classification tasks across four language models, four backdoor attack algorithms, and two different architectures of teacher models. Experimental results indicate success rates close to 100% for backdoor attacks targeting PEFT.
title Breaking PEFT Limitations: Leveraging Weak-to-Strong Knowledge Transfer for Backdoor Attacks in LLMs
topic Cryptography and Security
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2409.17946