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

Recently, adversarial attacks can be applied to the physical world, causingpractical issues to various Convolutional Neural Networks (CNNs) poweredapplications. Most existing physical adversarial attack defense works onlyfocus on eliminating explicit perturbation patterns from inputs, ignoringinterpretation to CNN's intrinsic vulnerability. Therefore, they lack theexpected versatility to different attacks and thereby depend on considerabledata processing costs. In this paper, we propose LanCe -- a comprehensive andlightweight CNN defense methodology against different physical adversarialattacks. By interpreting CNN's vulnerability, we find that non-semanticadversarial perturbations can activate CNN with significantly abnormalactivations and even overwhelm other semantic input patterns' activations. Weimprove the CNN recognition process by adding a self-verification stage todetect the potential adversarial input with only one CNN inference cost. Basedon the detection result, we further propose a data recovery methodology todefend the physical adversarial attacks. We apply such defense methodology intoboth image and audio CNN recognition scenarios and analyze the computationalcomplexity for each scenario, respectively. Experiments show that ourmethodology can achieve an average 91% successful rate for attack detection and89% accuracy recovery. Moreover, it is at most 3x faster compared with thestate-of-the-art defense methods, making it feasible to resource-constrainedembedded systems, such as mobile devices.

LanCe: A Comprehensive and Lightweight CNN Defense Methodology against Physical Adversarial Attacks on Embedded Multimedia Applications