Diffusion-Based Purification

DiffCAP: Diffusion-based Cumulative Adversarial Purification for Vision Language Models

This paper introduces DiffCAP, a novel diffusion-based purification strategy that can effectively neutralize adversarial corruptions in VLMs.

Jia Fu, Yongtao Wu, Yihang Chen, Kunyu Peng, Xiao Zhang, Volkan Cevher, Sepideh Pashami, Anders Holst