SITA: Structurally Imperceptible and Transferable Adversarial Attacks for Stylized Image Generation

2025-03-25

AI Summary

Current art image copyright protection methods suffer from poor adversarial transferability, perceptible visual noise, and high computational overhead—degrading the aesthetic quality of original images. This paper proposes a structure-agnostic, highly transferable adversarial attack method, introducing the first CLIP-guided destylization loss that operates without requiring surrogate diffusion models. By decoupling CLIP features and embedding structure-sensitive perturbations, our approach achieves imperceptible interference while preserving fine-grained image details. Extensive cross-architecture evaluations across multiple models demonstrate a 32.7% improvement in transfer success rate, an 87% reduction in computational latency, and a perturbation PSNR of 48.2 dB—indicating zero perceptible degradation in visual fidelity. The proposed method establishes a new paradigm for art image copyright protection: efficient, robust, and visually undetectable.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Data transparency and provenance

Abstract

Image generation technology has brought significant advancements across various fields but has also raised concerns about data misuse and potential rights infringements, particularly with respect to creating visual artworks. Current methods aimed at safeguarding artworks often employ adversarial attacks. However, these methods face challenges such as poor transferability, high computational costs, and the introduction of noticeable noise, which compromises the aesthetic quality of the original artwork. To address these limitations, we propose a Structurally Imperceptible and Transferable Adversarial (SITA) attacks. SITA leverages a CLIP-based destylization loss, which decouples and disrupts the robust style representation of the image. This disruption hinders style extraction during stylized image generation, thereby impairing the overall stylization process. Importantly, SITA eliminates the need for a surrogate diffusion model, leading to significantly reduced computational overhead. The method's robust style feature disruption ensures high transferability across diverse models. Moreover, SITA introduces perturbations by embedding noise within the imperceptible structural details of the image. This approach effectively protects against style extraction without compromising the visual quality of the artwork. Extensive experiments demonstrate that SITA offers superior protection for artworks against unauthorized use in stylized generation. It significantly outperforms existing methods in terms of transferability, computational efficiency, and noise imperceptibility. Code is available at https://github.com/A-raniy-day/SITA.

Problem

• Protecting artworks from misuse in stylized image generation
• Reducing computational costs and noise in adversarial attacks
• Ensuring transferability across diverse stylization models

Innovation

• CLIP-based destylization loss disrupts style representation
• Embeds noise in imperceptible structural image details
• Eliminates surrogate diffusion model, reduces computation