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/

Zendy Plus

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

Zendy Tools

Zendy Open

Deep neural networks (DNNs) are vulnerable to adversarial attack despitetheir tremendous success in many AI fields. Adversarial attack is a method thatcauses the intended misclassfication by adding imperceptible perturbations tolegitimate inputs. Researchers have developed numerous types of adversarialattack methods. However, from the perspective of practical deployment, thesemethods suffer from several drawbacks such as long attack generating time, highmemory cost, insufficient robustness and low transferability. We propose aContent-aware Adversarial Attack Generator (CAG) to achieve real-time,low-cost, enhanced-robustness and high-transferability adversarial attack.First, as a type of generative model-based attack, CAG shows significantspeedup (at least 500 times) in generating adversarial examples compared to thestate-of-the-art attacks such as PGD and C\&W. CAG only needs a singlegenerative model to perform targeted attack to any targeted class. Because CAGencodes the label information into a trainable embedding layer, it differs fromprior generative model-based adversarial attacks that use $n$ different copiesof generative models for $n$ different targeted classes. As a result, CAGsignificantly reduces the required memory cost for generating adversarialexamples. CAG can generate adversarial perturbations that focus on the criticalareas of input by integrating the class activation maps information in thetraining process, and hence improve the robustness of CAG attack against thestate-of-art adversarial defenses. In addition, CAG exhibits hightransferability across different DNN classifier models in black-box attackscenario by introducing random dropout in the process of generatingperturbations. Extensive experiments on different datasets and DNN models haveverified the real-time, low-cost, enhanced-robustness, and high-transferabilitybenefits of CAG.

CAG: A Real-time Low-cost Enhanced-robustness High-transferability Content-aware Adversarial Attack Generator