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

TL;DR

REFORGE is a black-box adversarial framework that tests the robustness of image generation model unlearning methods, revealing vulnerabilities and emphasizing the need for more robust unlearning techniques.

Contribution

This paper introduces REFORGE, a novel multi-modal attack framework that evaluates the robustness of image generation model unlearning in black-box settings.

Findings

REFORGE significantly increases attack success rates.

It achieves stronger semantic alignment and higher efficiency.

Vulnerabilities in current unlearning methods are exposed.

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.