Back to Search Start Over

Cross-Modal Safety Alignment: Is textual unlearning all you need?

Authors :
Chakraborty, Trishna
Shayegani, Erfan
Cai, Zikui
Abu-Ghazaleh, Nael
Asif, M. Salman
Dong, Yue
Roy-Chowdhury, Amit K.
Song, Chengyu
Publication Year :
2024

Abstract

Recent studies reveal that integrating new modalities into Large Language Models (LLMs), such as Vision-Language Models (VLMs), creates a new attack surface that bypasses existing safety training techniques like Supervised Fine-tuning (SFT) and Reinforcement Learning with Human Feedback (RLHF). While further SFT and RLHF-based safety training can be conducted in multi-modal settings, collecting multi-modal training datasets poses a significant challenge. Inspired by the structural design of recent multi-modal models, where, regardless of the combination of input modalities, all inputs are ultimately fused into the language space, we aim to explore whether unlearning solely in the textual domain can be effective for cross-modality safety alignment. Our evaluation across six datasets empirically demonstrates the transferability -- textual unlearning in VLMs significantly reduces the Attack Success Rate (ASR) to less than 8\% and in some cases, even as low as nearly 2\% for both text-based and vision-text-based attacks, alongside preserving the utility. Moreover, our experiments show that unlearning with a multi-modal dataset offers no potential benefits but incurs significantly increased computational demands, possibly up to 6 times higher.

Details

Database :
arXiv
Publication Type :
Report
Accession number :
edsarx.2406.02575
Document Type :
Working Paper