eduzhai > Applied Sciences > Engineering >

Adversarial Attacks against Face Recognition A Comprehensive Study

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 30 pages

Abstract: Face recognition (FR) systems have demonstrated outstanding verificationperformance, suggesting suitability for real-world applications ranging fromphoto tagging in social media to automated border control (ABC). In an advancedFR system with deep learning-based architecture, however, promoting therecognition efficiency alone is not sufficient, and the system should alsowithstand potential kinds of attacks designed to target its proficiency. Recentstudies show that (deep) FR systems exhibit an intriguing vulnerability toimperceptible or perceptible but natural-looking adversarial input images thatdrive the model to incorrect output predictions. In this article, we present acomprehensive survey on adversarial attacks against FR systems and elaborate onthe competence of new countermeasures against them. Further, we propose ataxonomy of existing attack and defense methods based on different criteria. Wecompare attack methods on the orientation and attributes and defense approacheson the category. Finally, we explore the challenges and potential researchdirection.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×