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-language pre-training (VLP) models demonstrate impressive abilities inprocessing both images and text. However, they are vulnerable to multi-modaladversarial examples (AEs). Investigating the generation ofhigh-transferability adversarial examples is crucial for uncovering VLP models'vulnerabilities in practical scenarios. Recent works have indicated thatleveraging data augmentation and image-text modal interactions can enhance thetransferability of adversarial examples for VLP models significantly. However,they do not consider the optimal alignment problem between dataaugmentedimage-text pairs. This oversight leads to adversarial examples that are overlytailored to the source model, thus limiting improvements in transferability. Inour research, we first explore the interplay between image sets producedthrough data augmentation and their corresponding text sets. We find thataugmented image samples can align optimally with certain texts while exhibitingless relevance to others. Motivated by this, we propose an OptimalTransport-based Adversarial Attack, dubbed OT-Attack. The proposed methodformulates the features of image and text sets as two distinct distributionsand employs optimal transport theory to determine the most efficient mappingbetween them. This optimal mapping informs our generation of adversarialexamples to effectively counteract the overfitting issues. Extensiveexperiments across various network architectures and datasets in image-textmatching tasks reveal that our OT-Attack outperforms existing state-of-the-artmethods in terms of adversarial transferability.

OT-Attack: Enhancing Adversarial Transferability of Vision-Language Models via Optimal Transport Optimization