Kernel-based adversarial attacks and defenses on support vector classification

作     者：Wanman Li Xiaozhang Liu Anli Yan Jie Yang Wanman Li;Xiaozhang Liu;Anli Yan;Jie Yang

作者机构：School of Computer Science and TechnologyHainan UniversityHaikou570228China 

出 版 物：《Digital Communications and Networks》 (数字通信与网络（英文版）)

年 卷 期：2022年第8卷第4期

页      面：492-497页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the National Natural Science Foundation of China under Grant No.61966011 

主　　题：Adversarial machine learning Support vector machines Evasion attack Vulnerability function Kernel optimization 

摘      要：While malicious samples are widely found in many application fields of machine learning,suitable countermeasures have been investigated in the field of adversarial machine *** to the importance and popularity of Support Vector Machines(SVMs),we first describe the evasion attack against SVM classification and then propose a defense strategy in this *** evasion attack utilizes the classification surface of SVM to iteratively find the minimal perturbations that mislead the nonlinear ***,we propose what is called a vulnerability function to measure the vulnerability of the SVM *** this vulnerability function,we put forward an effective defense strategy based on the kernel optimization of SVMs with Gaussian kernel against the evasion *** defense method is verified to be very effective on the benchmark datasets,and the SVM classifier becomes more robust after using our kernel optimization scheme.