基于PSO-TSA模型的网络安全态势要素识别研究  

Research on Recognition of Network Security Situation Elements Based on PSO-TSA Model

作　　者：张克君 郑炜 于新颖 王航宇 王志强 ZHANG Kejun;ZHENG Wei;YU Xinying;WANG Hangyu;WANG Zhiqiang(Department of Cyberspace Security,Beijing Electronic Science and Technology Institute,Beijing 100071,China;College of Cyberspace Security,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]北京电子科技学院网络空间安全系,北京100071 [2]北京邮电大学网络空间安全学院,北京100876

出　　处：《湖南大学学报（自然科学版）》2022年第4期119-127,共9页Journal of Hunan University:Natural Sciences

基　　金：北京高校高精尖学科建设项目(20210086Z0401);国家重点研发计划网络空间安全重大专项课题资助(2018YFB0803601)。

摘　　要：针对网络安全态势感知技术中态势要素提取的质量与效率较低的问题,提出了融合粒子群(Particle Swarm Optimization,PSO)和模拟退火(Simulated Annealing,SA)的态势要素识别模型PSO-TSA.在位置更新模块,利用Metropolis准则对PSO算法中的个体极值和全局极值进行退火优化,增加粒子的选择性,提高态势要素提取质量.在参数优化模块,利用Metropo⁃lis准则优化PSO算法中的参数,并对参数优化过程和粒子适应度同时进行评价,避免算法陷入局部最优,提高态势要素识别效率.按照目前网络状态的实际需求,选择了37个网络安全数据字段,搭建了小型网络环境,以获取更加真实的网络安全数据集SDS-W.在开放网络安全数据集和获取的SDS-W数据集上分别进行态势要素识别实验,实验证明,PSO-TSA在时间成本保持不变甚至更少的基础上,态势要素识别的精确度平均提升了5%~7%.Given the low quality and efficiency of situation element extraction in network security situation awareness techniques,this paper proposes a situation element identification model incorporating particle swarm opti⁃mization and simulated annealing(PSO-TSA).In the position update module,the Metropolis criterion is utilized to optimize the individual and global extremum in the PSO algorithm to increase the selectivity of the particles and im⁃prove the quality of the situation elements extraction.In the parameter optimization module,the parameters in the PSO algorithm are optimized using the Metropolis criterion,and the parameter optimization process and particle fit⁃ness are evaluated simultaneously to rid the local optimum and improve the efficiency of the situation element recog⁃nition.Due to the actual needs of the current network state,this paper selects 37 network security data fields and es⁃tablishes a small network environment to obtain a more realistic network security dataset SDS-W.This paper con⁃ducts experiments of the situation element recognition on the open cybersecurity dataset and the SDS-W,respec⁃tively.Experiments show that PSO-TSA improves the accuracy of situation element recognition by an average of 5%to 7%while the time cost remains the same or even less.

关 键 词：网络安全态势感知 态势要素识别 粒子群算法 模拟退火算法 

分 类 号：TN915.08[电子电信—通信与信息系统]