Effective and Efficient Adversarial Detection for Vision-Language Models via A Single Vector

Youcheng Huang, Fengbin Zhu, Jingkun Tang, Pan Zhou, Wenqiang Lei +2 more
10/30/2024
cs.CVcs.CLcs.CR

Official code: mob-scu/RADAR-NEARSIDE (11 stars, last commit 2024-10-30).

Abstract

Visual Language Models (VLMs) are vulnerable to adversarial attacks, especially those from adversarial images, which is however under-explored in literature. To facilitate research on this critical safety problem, we first construct a new laRge-scale Adervsarial images dataset with Diverse hArmful Responses (RADAR), given that existing datasets are either small-scale or only contain limited types of harmful responses. With the new RADAR dataset, we further develop a novel and effective iN-time Embedding-based AdveRSarial Image DEtection (NEARSIDE) method, which exploits a single vector that distilled from the hidden states of VLMs, which we call the attacking direction, to achieve the detection of adversarial images against benign ones in the input. Extensive experiments with two victim VLMs, LLaVA and MiniGPT-4, well demonstrate the effectiveness, efficiency, and cross-model transferrability of our proposed method. Our code is available at https://github.com/mob-scu/RADAR-NEARSIDE

View on arXivView PDF

Code Implementations(1)

mob-scu/RADAR-NEARSIDEOfficial100%
110Shell, PythonOct 10, 20241 years ago

Similar Papers

MirrorCheck: Efficient Adversarial Defense for Vision-Language Models

89% match2024

Towards Adversarially Robust Vision-Language Models: Insights from Design Choices and Prompt Formatting Techniques

89% match2024

Adversarial Defense in Vision-Language Models: An Overview

88% match2026

Doubly-Universal Adversarial Perturbations: Deceiving Vision-Language Models Across Both Images and Text with a Single Perturbation

88% match2024

On Evaluating Adversarial Robustness of Large Vision-Language Models

88% match1 repo2262023

A Hybrid Defense Strategy for Boosting Adversarial Robustness in Vision-Language Models

87% match2024

Cite this paper

@article{huang2024effective,
  title  = {Effective and Efficient Adversarial Detection for Vision-Language Models via A Single Vector},
  author = {Youcheng Huang and Fengbin Zhu and Jingkun Tang and Pan Zhou and Wenqiang Lei and Jiancheng Lv and Tat-Seng Chua},
  year   = {2024},
  eprint = {2410.22888},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2410.22888v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2410.22888.svg)](https://codeofpaper.com/paper/2410.22888)

Discussion