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

Neural networks are known to be susceptible to adversarial samples: smallvariations of natural examples crafted to deliberately mislead the models.While they can be easily generated using gradient-based techniques in digitaland physical scenarios, they often differ greatly from the actual datadistribution of natural images, resulting in a trade-off between strength andstealthiness. In this paper, we propose a novel framework dubbedDiffusion-Based Projected Gradient Descent (Diff-PGD) for generating realisticadversarial samples. By exploiting a gradient guided by a diffusion model,Diff-PGD ensures that adversarial samples remain close to the original datadistribution while maintaining their effectiveness. Moreover, our framework canbe easily customized for specific tasks such as digital attacks, physical-worldattacks, and style-based attacks. Compared with existing methods for generatingnatural-style adversarial samples, our framework enables the separation ofoptimizing adversarial loss from other surrogate losses (e.g.,content/smoothness/style loss), making it more stable and controllable.Finally, we demonstrate that the samples generated using Diff-PGD have bettertransferability and anti-purification power than traditional gradient-basedmethods. Code will be released in https://github.com/xavihart/Diff-PGD

Diffusion-Based Adversarial Sample Generation for Improved Stealthiness and Controllability