基于混合差分演化的网络入侵检测算法-《郑州大学学报(工学版)》

文章信息/Info

Title:: Network intrusion detection algorithm based on hybrid differential evolution algorithm

作者:: 王耀光¹; 陈伟权¹; 吴镇邦¹; 秦勇²; 黄翰³; 1.广东省东莞市质量监督检测中心,广东东莞,523000;2.东莞理工学院计算机学院,广东东莞,523000;3.华南理工大学软件学院,广东广州,510006

Author(s):: Wang Yaoguang1; Chen Weiquan1; Wu Zhenbang1; Qin Yong2Huang Han3; 1. Quality Supervision and Testing Center of Dongguan City, Guangdong Province, Dongguan, Guangdong, 523000; 2. School of Computer Science, Dongguan Institute of Technology, Dongguan, Guangdong 523000; 3. Software College, South China University of Technology, Guangzhou, Guangdong 510006

Keywords:: network intrusion detection; stability test; hybrid differential evolution; least squares support vector machine

摘要:: 基于机器学习方法的入侵检测算法是目前网络设备检测领域的研究热点。网络入侵检测源数据的多样性是影响机器学习方法在该领域实际应用性能的主要因素。研究通过设计多扰动向量混合差分演化算法，稳定地优化了最小二乘支持向量机模型的关键参数；在不增加测试集检测计算复杂性的前提下，通过最优化参数的方式，提高了最小二乘支持向量机算法入侵检测的精度和稳定性。KDD Cup 09测试集的仿真实验结果显示，所提出的基于混合差分演化的网络入侵监测算法比目前多种同类算法有着更好的平均性能。

Abstract:: Intrusion detection algorithm based on machine learning method is one of research hotspot in the field of network equiment testing. In the face of the real-worid application requirement, machine learning methods should be further optimized to achieve accurate and stable detection effect. The study optimize steadily several key parameters of least squares support vector machine (SVM) by designing a hybird differential evolution algorithm with disturbance vector and improved the intrusion detection accuracy and stability of least squares support vector machine (SVM) algorithm by means of adaptive parameter turning. The experimental results in KDD Cup 09 test set showed that, the proposed network intrusion detedtion algorithm based on hybird dieeerential evolution algorithm had better performance on average than many similar algorithm at present.