English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Despite the success of diffusion-based customization methods on visualcontent creation, increasing concerns have been raised about such techniquesfrom both privacy and political perspectives. To tackle this issue, severalanti-customization methods have been proposed in very recent months,predominantly grounded in adversarial attacks. Unfortunately, most of thesemethods adopt straightforward designs, such as end-to-end optimization with afocus on adversarially maximizing the original training loss, therebyneglecting nuanced internal properties intrinsic to the diffusion model, andeven leading to ineffective optimization in some diffusion time steps. In thispaper, we strive to bridge this gap by undertaking a comprehensive explorationof these inherent properties, to boost the performance of currentanti-customization approaches. Two aspects of properties are investigated: 1)We examine the relationship between time step selection and the model'sperception in the frequency domain of images and find that lower time steps cangive much more contributions to adversarial noises. This inspires us to proposean adaptive greedy search for optimal time steps that seamlessly integrateswith existing anti-customization methods. 2) We scrutinize the roles offeatures at different layers during denoising and devise a sophisticatedfeature-based optimization framework for anti-customization. Experiments onfacial benchmarks demonstrate that our approach significantly increasesidentity disruption, thereby enhancing user privacy and security.

SimAC: A Simple Anti-Customization Method against Text-to-Image Synthesis of Diffusion Models