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

Vision transformers (ViTs) have been successfully deployed in a variety ofcomputer vision tasks, but they are still vulnerable to adversarial samples.Transfer-based attacks use a local model to generate adversarial samples anddirectly transfer them to attack a target black-box model. The high efficiencyof transfer-based attacks makes it a severe security threat to ViT-basedapplications. Therefore, it is vital to design effective transfer-based attacksto identify the deficiencies of ViTs beforehand in security-sensitivescenarios. Existing efforts generally focus on regularizing the input gradientsto stabilize the updated direction of adversarial samples. However, thevariance of the back-propagated gradients in intermediate blocks of ViTs maystill be large, which may make the generated adversarial samples focus on somemodel-specific features and get stuck in poor local optima. To overcome theshortcomings of existing approaches, we propose the Token GradientRegularization (TGR) method. According to the structural characteristics ofViTs, TGR reduces the variance of the back-propagated gradient in each internalblock of ViTs in a token-wise manner and utilizes the regularized gradient togenerate adversarial samples. Extensive experiments on attacking both ViTs andCNNs confirm the superiority of our approach. Notably, compared to thestate-of-the-art transfer-based attacks, our TGR offers a performanceimprovement of 8.8% on average.

Transferable Adversarial Attacks on Vision Transformers with Token Gradient Regularization