基于SVM和DS证据理论的网络攻击检测研究  

Research of Network Attacks Detection Based on SVM and DS Evidence Theory

作　　者：白龙[1] 孙强[2] 

机构地区：[1]牡丹江师范学院物理与电子工程学院,黑龙江牡丹江157011 [2]牡丹江师范学院工学院,黑龙江牡丹江157011

出　　处：《科技通报》2017年第6期196-200,共5页Bulletin of Science and Technology

基　　金：黑龙江省自然科学基金(QC2013C067);牡丹江师范学院省级重点创新预研项目(SY201218)

摘　　要：为了解决传统网络攻击检测算法存在的检测准确率低等问题,提出了一种基于SVM和DS证据理论的网络攻击检测算法。该算法首先利用样本主特征进行检测,当数据出现模糊分类时,利用辅助特征和DS证据理论对数据的隶属度进行重新划分,有效改善了经典SVM算法的模糊分类问题。仿真结果表明,带有修正机制的SVM算法对于网络攻击检测,尤其是攻击类型未知的数据,检测准确率明显提高,平均到了95%以上,算法的整体性能表现优良。To solve the problems of traditional network attacks detection algorithms, such as poor detected accuracy, a new detection algorithm based on SVM and DS evidence theory is proposed. This algorithm performs detection using principal features first, and then the assist features and DS evidence theory are adopted to recalculate the membership of the data classified uncertainly, which effectively overcomes the fuzzy classification problem of the classical SVM algorithm. Simulation results show that the detected accuracy of network attacks is significantly improved by the modified SVM, especially for the unknown types of attacks. The average value is more than 95%, and the overall performance of this algorithm is so good.

关 键 词：网络攻击检测 SVM 证据理论 DS合成规则 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]