T2IShield: Defending Against Backdoors on Text-to-Image Diffusion Models

  • Zhongqi Wang
  • , Jie Zhang*
  • , Shiguang Shan
  • , Xilin Chen
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

3 Citations (Scopus)

Abstract

While text-to-image diffusion models demonstrate impressive generation capabilities, they also exhibit vulnerability to backdoor attacks, which involve the manipulation of model outputs through malicious triggers. In this paper, for the first time, we propose a comprehensive defense method named T2IShield to detect, localize, and mitigate such attacks. Specifically, we find the “Assimilation Phenomenon” on the cross-attention maps caused by the backdoor trigger. Based on this key insight, we propose two effective backdoor detection methods: Frobenius Norm Threshold Truncation and Covariance Discriminant Analysis. Besides, we introduce a binary-search approach to localize the trigger within a backdoor sample and assess the efficacy of existing concept editing methods in mitigating backdoor attacks. Empirical evaluations on two advanced backdoor attack scenarios show the effectiveness of our proposed defense method. For backdoor sample detection, T2IShield achieves a detection F1 score of 88.9% with low computational cost. Furthermore, T2IShield achieves a localization F1 score of 86.4% and invalidates 99% poisoned samples. Codes are released at https://github.com/Robin-WZQ/T2IShield.

Original languageEnglish
Title of host publicationComputer Vision – ECCV 2024 - 18th European Conference, Proceedings
EditorsAleš Leonardis, Elisa Ricci, Stefan Roth, Olga Russakovsky, Torsten Sattler, Gül Varol
PublisherSpringer Science and Business Media Deutschland GmbH
Pages107-124
Number of pages18
ISBN (Print)9783031730122
DOIs
Publication statusPublished - 2025
Externally publishedYes
Event18th European Conference on Computer Vision, ECCV 2024 - Milan, Italy
Duration: 29 Sept 20244 Oct 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume15143 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference18th European Conference on Computer Vision, ECCV 2024
Country/TerritoryItaly
CityMilan
Period29/09/244/10/24

Keywords

  • Backdoor defence
  • Backdoor detection
  • Backdoor mitigation
  • Text-to-image diffusion models

Fingerprint

Dive into the research topics of 'T2IShield: Defending Against Backdoors on Text-to-Image Diffusion Models'. Together they form a unique fingerprint.

Cite this