Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

REFORGE: Multi-modal Attacks Reveal Vulnerable Concept Unlearning in Image Generation Models

Yong Zou, Haoran Li, Fanxiao Li, Shenyang Wei, Yunyun Dong, Li Tang, Wei Zhou, Renyang Liu

Abstract

Recent progress in image generation models (IGMs) enables high-fidelity content creation but also amplifies risks, including the reproduction of copyrighted content and the generation of offensive content. Image Generation Model Unlearning (IGMU) mitigates these risks by removing harmful concepts without full retraining. Despite growing attention, the robustness under adversarial inputs, particularly image-side threats in black-box settings, remains underexplored. To bridge this gap, we present REFORGE, a black-box red-teaming framework that evaluates IGMU robustness via adversarial image prompts. REFORGE initializes stroke-based images and optimizes perturbations with a cross-attention-guided masking strategy that allocates noise to concept-relevant regions, balancing attack efficacy and visual fidelity. Extensive experiments across representative unlearning tasks and defenses demonstrate that REFORGE significantly improves attack success rate while achieving stronger semantic alignment and higher efficiency than involved baselines. These results expose persistent vulnerabilities in current IGMU methods and highlight the need for robustness-aware unlearning against multi-modal adversarial attacks. Our code is at: https://github.com/Imfatnoily/REFORGE.

REFORGE: Multi-modal Attacks Reveal Vulnerable Concept Unlearning in Image Generation Models

Abstract

Recent progress in image generation models (IGMs) enables high-fidelity content creation but also amplifies risks, including the reproduction of copyrighted content and the generation of offensive content. Image Generation Model Unlearning (IGMU) mitigates these risks by removing harmful concepts without full retraining. Despite growing attention, the robustness under adversarial inputs, particularly image-side threats in black-box settings, remains underexplored. To bridge this gap, we present REFORGE, a black-box red-teaming framework that evaluates IGMU robustness via adversarial image prompts. REFORGE initializes stroke-based images and optimizes perturbations with a cross-attention-guided masking strategy that allocates noise to concept-relevant regions, balancing attack efficacy and visual fidelity. Extensive experiments across representative unlearning tasks and defenses demonstrate that REFORGE significantly improves attack success rate while achieving stronger semantic alignment and higher efficiency than involved baselines. These results expose persistent vulnerabilities in current IGMU methods and highlight the need for robustness-aware unlearning against multi-modal adversarial attacks. Our code is at: https://github.com/Imfatnoily/REFORGE.
Paper Structure (22 sections, 6 equations, 3 figures, 2 tables, 1 algorithm)

This paper contains 22 sections, 6 equations, 3 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Image Generation Model Unlearning
  4. Red Teaming for Image Generation Model Unlearning
  5. Methodology
  6. Threat Model
  7. Overview
  8. Initialization of Adversarial Sample
  9. Mask Construction via Cross-Attention
  10. Latent-Alignment Optimization
  11. Red-Teaming Evaluation
  12. Experiments
  13. Settings
  14. Attack Performance
  15. Semantic Alignment
  16. ...and 7 more sections

Figures (3)

  • Figure 1: Given that an unlearned IGM has undergone a concept-unlearning procedure (e.g., removal of the Van Gogh style), our adversarial image prompt $P_{adv}$ combined with the prompt $P_{text}$ can still bypass the unlearning mechanism, causing the erased style to re-emerge in the generated image $I^{*}$.
  • Figure 2: Overview of the ReForge framework. Sensitive parts are covered by .
  • Figure 3: Ablation of key parameters: ASR (%) vs. (a) reference images, (b) timesteps, (c) layers, and (d) losses.