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 neural networks (DNNs) are vulnerable to adversarial examples, which maylead to catastrophe in security-critical domains. Numerous detection methodsare proposed to characterize the feature uniqueness of adversarial examples, orto distinguish DNN's behavior activated by the adversarial examples. Detectionsbased on features cannot handle adversarial examples with large perturbations.Besides, they require a large amount of specific adversarial examples. Anothermainstream, model-based detections, which characterize input properties bymodel behaviors, suffer from heavy computation cost. To address the issues, weintroduce the concept of local gradient, and reveal that adversarial exampleshave a quite larger bound of local gradient than the benign ones. Inspired bythe observation, we leverage local gradient for detecting adversarial examples,and propose a general framework AdvCheck. Specifically, by calculating thelocal gradient from a few benign examples and noise-added misclassifiedexamples to train a detector, adversarial examples and even misclassifiednatural inputs can be precisely distinguished from benign ones. Throughextensive experiments, we have validated the AdvCheck's superior performance tothe state-of-the-art (SOTA) baselines, with detection rate ($\sim \times 1.2$)on general adversarial attacks and ($\sim \times 1.4$) on misclassified naturalinputs on average, with average 1/500 time cost. We also provide interpretableresults for successful detection.

AdvCheck: Characterizing Adversarial Examples via Local Gradient Checking