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

Adversarial attacks can make deep neural network (DNN) models predictincorrect output labels, such as misclassified traffic signs, for autonomousvehicle (AV) perception modules. Resilience against adversarial attacks canhelp AVs navigate safely on the road by avoiding misclassication of signs orobjects. This DNN-based study develops a resilient traffic sign classifier forAVs that uses a hybrid defense method. We use transfer learning to retrain theInception-V3 and Resnet-152 models as traffic sign classifiers. This methodalso utilizes a combination of three different strategies: random filtering,ensembling, and local feature mapping. We use the random cropping and resizingtechnique for random filtering, plurality voting as ensembling strategy and anoptical character recognition model as a local feature mapper. This DNN-basedhybrid defense method has been tested for the no attack scenario and againstwell-known untargeted adversarial attacks (e.g., Projected Gradient Descent orPGD, Fast Gradient Sign Method or FGSM, Momentum Iterative Method or MIMattack, and Carlini and Wagner or C&W). We find that our hybrid defense methodachieves 99% average traffic sign classification accuracy for the no attackscenario and 88% average traffic sign classification accuracy for all attackscenarios. Moreover, the hybrid defense method, presented in this study,improves the accuracy for traffic sign classification compared to thetraditional defense methods (i.e., JPEG filtering, feature squeezing, binaryfiltering, and random filtering) up to 6%, 50%, and 55% for FGSM, MIM, and PGDattacks, respectively.

A Hybrid Defense Method against Adversarial Attacks on Traffic Sign Classifiers in Autonomous Vehicles