Unlearning Backdoor Threats: Enhancing Backdoor Defense in Multimodal Contrastive Learning via Local Token Unlearning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liang, Siyuan, Liu, Kuanrong, Gong, Jiajun, Liang, Jiawei, Xun, Yuan, Chang, Ee-Chien, Cao, Xiaochun
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917621529575424
author Liang, Siyuan
Liu, Kuanrong
Gong, Jiajun
Liang, Jiawei
Xun, Yuan
Chang, Ee-Chien
Cao, Xiaochun
author_facet Liang, Siyuan
Liu, Kuanrong
Gong, Jiajun
Liang, Jiawei
Xun, Yuan
Chang, Ee-Chien
Cao, Xiaochun
contents Multimodal contrastive learning has emerged as a powerful paradigm for building high-quality features using the complementary strengths of various data modalities. However, the open nature of such systems inadvertently increases the possibility of backdoor attacks. These attacks subtly embed malicious behaviors within the model during training, which can be activated by specific triggers in the inference phase, posing significant security risks. Despite existing countermeasures through fine-tuning that reduce the adverse impacts of such attacks, these defenses often degrade the clean accuracy and necessitate the construction of extensive clean training pairs. In this paper, we explore the possibility of a less-cost defense from the perspective of model unlearning, that is, whether the model can be made to quickly \textbf{u}nlearn \textbf{b}ackdoor \textbf{t}hreats (UBT) by constructing a small set of poisoned samples. Specifically, we strengthen the backdoor shortcuts to discover suspicious samples through overfitting training prioritized by weak similarity samples. Building on the initial identification of suspicious samples, we introduce an innovative token-based localized forgetting training regime. This technique specifically targets the poisoned aspects of the model, applying a focused effort to unlearn the backdoor associations and trying not to damage the integrity of the overall model. Experimental results show that our method not only ensures a minimal success rate for attacks, but also preserves the model's high clean accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2403_16257
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Unlearning Backdoor Threats: Enhancing Backdoor Defense in Multimodal Contrastive Learning via Local Token Unlearning
Liang, Siyuan
Liu, Kuanrong
Gong, Jiajun
Liang, Jiawei
Xun, Yuan
Chang, Ee-Chien
Cao, Xiaochun
Computer Vision and Pattern Recognition
Multimodal contrastive learning has emerged as a powerful paradigm for building high-quality features using the complementary strengths of various data modalities. However, the open nature of such systems inadvertently increases the possibility of backdoor attacks. These attacks subtly embed malicious behaviors within the model during training, which can be activated by specific triggers in the inference phase, posing significant security risks. Despite existing countermeasures through fine-tuning that reduce the adverse impacts of such attacks, these defenses often degrade the clean accuracy and necessitate the construction of extensive clean training pairs. In this paper, we explore the possibility of a less-cost defense from the perspective of model unlearning, that is, whether the model can be made to quickly \textbf{u}nlearn \textbf{b}ackdoor \textbf{t}hreats (UBT) by constructing a small set of poisoned samples. Specifically, we strengthen the backdoor shortcuts to discover suspicious samples through overfitting training prioritized by weak similarity samples. Building on the initial identification of suspicious samples, we introduce an innovative token-based localized forgetting training regime. This technique specifically targets the poisoned aspects of the model, applying a focused effort to unlearn the backdoor associations and trying not to damage the integrity of the overall model. Experimental results show that our method not only ensures a minimal success rate for attacks, but also preserves the model's high clean accuracy.
title Unlearning Backdoor Threats: Enhancing Backdoor Defense in Multimodal Contrastive Learning via Local Token Unlearning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.16257