Quantifying the resilience of machine learning classifiers used for cyber security

作者：

Highlights：

• Quantifying machine learning classifiers’ resilience to adversarial manipulations.

• Formal model for evaluating attacker's budget and the feature manipulation cost.

• Present two adversary aware feature selection using budget and manipulation cost.

• Demonstrate our approach using real life malware and benign executable analysis.

摘要

•Quantifying machine learning classifiers’ resilience to adversarial manipulations.•Formal model for evaluating attacker's budget and the feature manipulation cost.•Present two adversary aware feature selection using budget and manipulation cost.•Demonstrate our approach using real life malware and benign executable analysis.

论文关键词：Adversarial Learning,Classifier Resilience,Cyber Security

论文评审过程：Received 24 June 2017, Revised 23 September 2017, Accepted 24 September 2017, Available online 29 September 2017, Version of Record 6 October 2017.

论文官网地址：https://doi.org/10.1016/j.eswa.2017.09.053