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

The widespread application of deep neural network (DNN) techniques is beingchallenged by adversarial examples, the legitimate input added withimperceptible and well-designed perturbations that can fool DNNs easily in theDNN testing/deploying stage. Previous adversarial example generation algorithmsfor adversarial white-box attacks used Jacobian gradient information to addperturbations. This information is too imprecise and inexplicit, which willcause unnecessary perturbations when generating adversarial examples. Thispaper aims to address this issue. We first propose to apply a more informativeand distilled gradient information, namely integrated gradient, to generateadversarial examples. To further make the perturbations more imperceptible, wepropose to employ the restriction combination of $L_0$ and $L_1/L_2$ secondly,which can restrict the total perturbations and perturbation pointssimultaneously. Meanwhile, to address the non-differentiable problem of $L_1$,we explore a proximal operation of $L_1$ thirdly. Based on these three works,we propose two Integrated gradient based White-box Adversarial examplegeneration algorithms (IWA): IFPA and IUA. IFPA is suitable for situationswhere there are a determined number of points to be perturbed. IUA is suitablefor situations where no perturbation point number is preset in order to obtainmore adversarial examples. We verify the effectiveness of the proposedalgorithms on both structured and unstructured datasets, and we compare themwith five baseline generation algorithms. The results show that our proposedalgorithms do craft adversarial examples with more imperceptible perturbationsand satisfactory crafting rate. $L_2$ restriction is more suitable forunstructured dataset and $L_1$ restriction performs better in structureddataset.

IWA: Integrated Gradient based White-box Attacks for Fooling Deep Neural Networks