基于特征再挑选的网络未知流量检测算法  

Network unknown traffic detection algorithm based on feature re-selection

作　　者：王忠勇[1] 孟杰[2] 王玮[1] 巩克现 刘宏华[3] WANG Zhong-yong;MENG Jie;WANG Wei;GONG Ke-xian;LIU Hong-hua(School of Electrical and Information Engineering,Zhengzhou University,Zhengzhou 450001,China;School of Cyber Science and Engineering,Zhengzhou University,Zhengzhou 450002,China;Twenty-seventh Research Institute,China Electronics Technology Group Corporation,Zhengzhou 450047,China)

机构地区：[1]郑州大学电气与信息工程学院,河南郑州450001 [2]郑州大学网络空间安全学院,河南郑州450002 [3]中国电子科技集团公司第二十七研究所,河南郑州450047

出　　处：《计算机工程与设计》2025年第1期60-66,共7页Computer Engineering and Design

基　　金：国家重点研发计划基金项目(2019QY0302);国家自然科学基金项目(61901417)。

摘　　要：为解决未知流量检测研究中不同种类流量因存在相同的结构或字段而导致检出率下降的问题,提出一种Open-SFSP(open-set selection feature and subspace projection)网络未知流量检测算法。在Open-MUSIC算法的基础上增加特征提取网络输出的特征维度,定义特征偏移距离与特征偏移数量以衡量特征的偏移程度,以偏移程度为指标挑选相较已知流量特征偏移程度大的特征完成后续未知流量检测步骤。实验结果表明,Open-SFSP算法相较Open-MUSIC算法在不同数据集上都表现出明显的效果提升,具有较高的准确性和可靠性。该算法为网络安全领域中的未知流量检测提供了一种有效的解决方案。To cope with the problem that the detection rate decreases due to the same structure or field of different types of network traffic in the research of unknown traffic detection,an Open-SFSP(open-set selection feature and subspace projection)algorithm for network unknown traffic detection was proposed.Based on the Open-MUSIC algorithm,the feature dimension of model output was increased,and the feature offset degree of the unknown traffic and known traffic was measured by defining the feature offset distance and offset number,and the features with greater offset degree than known traffic were selected to complete the subsequent unknown traffic detection.Numerical experiments verify that the proposed algorithm exhibits significant improvement compared with the Open-MUSIC algorithm on different data sets,showing higher accuracy and reliability.This algorithm can provide an effective solution for network unknown traffic detection in the field of network security.

关 键 词：未知流量检测 特征提取网络 特征偏移距离 特征偏移数量 特征偏移程度 特征再挑选 投影 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]