面向SVM的隐私保护方法研究进展

Research progress of privacy-preserving support vector machines

针对未来应用SVM进行数据挖掘所面临的信息安全问题,对隐私保护支持向量机分类规则挖掘方法进行研究,以提高支持向量机进行分类时的数据安全性,同时获得有效结果.分析了支持向量机分类方法的特点和可能面临的安全威胁;对国内外相关研究成果进行了归纳和梳理;重点从数据干扰和数据加密2个角度,给出了支持向量机隐私保护技术的最新研究进展;归纳出目前研究存在的问题和未来研究的趋势.指出了支持向量机隐私保护的研究方向:分布式环境下局部分类器融合隐私保护策略、更高效率的全同态加密方案、保护SVM分类规则的方案以及适用于大数据挖掘的隐私保护SVM技术.

To realize information security for future support vector machines （SVM）data mining,the privacy-preserving support vector machines （PPSVM） was investigated to obtain effective result.The characteristics of SVMclassifiers were analyzed to find the security hole.The latest literatures and related research were summarized. The recent progress of privacy-preserving support vector machines was presented based on data perturbation and data encryption.The future hot research directions of new privacy-preserving support vector machine technologies in distributed environment,more effective fully homomorphic encryption（FHE）schemes and privacy-preserving support vector machine technologies for big data mining were pointed out.