期刊
IEEE ACCESS
卷 10, 期 -, 页码 998-1019出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/ACCESS.2021.3138338
关键词
Adversarial machine learning; Training; Task analysis; Training data; Licenses; Image recognition; Security; Adversarial machine learning; adversarial examples; adversarial defense; black-box attack; security; deep learning
This paper provides a systematic survey and categorization of recent advances in adversarial machine learning black-box attacks. It summarizes 20 black-box attacks and introduces a new mathematical framework for evaluating attack results.
The field of adversarial machine learning has experienced a near exponential growth in the amount of papers being produced since 2018. This massive information output has yet to be properly processed and categorized. In this paper, we seek to help alleviate this problem by systematizing the recent advances in adversarial machine learning black-box attacks since 2019. Our survey summarizes and categorizes 20 recent black-box attacks. We also present a new analysis for understanding the attack success rate with respect to the adversarial model used in each paper. Overall, our paper surveys a wide body of literature to highlight recent attack developments and organizes them into four attack categories: score based attacks, decision based attacks, transfer attacks and non-traditional attacks. Further, we provide a new mathematical framework to show exactly how attack results can fairly be compared.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据