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

Deep learning models are vulnerable to adversarial examples crafted byapplying human-imperceptible perturbations on benign inputs. However, under theblack-box setting, most existing adversaries often have a poor transferabilityto attack other defense models. In this work, from the perspective of regardingthe adversarial example generation as an optimization process, we propose twonew methods to improve the transferability of adversarial examples, namelyNesterov Iterative Fast Gradient Sign Method (NI-FGSM) and Scale-Invariantattack Method (SIM). NI-FGSM aims to adapt Nesterov accelerated gradient intothe iterative attacks so as to effectively look ahead and improve thetransferability of adversarial examples. While SIM is based on our discovery onthe scale-invariant property of deep learning models, for which we leverage tooptimize the adversarial perturbations over the scale copies of the inputimages so as to avoid "overfitting" on the white-box model being attacked andgenerate more transferable adversarial examples. NI-FGSM and SIM can benaturally integrated to build a robust gradient-based attack to generate moretransferable adversarial examples against the defense models. Empirical resultson ImageNet dataset demonstrate that our attack methods exhibit highertransferability and achieve higher attack success rates than state-of-the-artgradient-based attacks.

Nesterov Accelerated Gradient and Scale Invariance for Adversarial Attacks