入侵检测与聚类分析  

Intrusion Detection And Clustering Analysis

作　　者：刘勇国[1] 彭军[2] 张伟[3] 李学明[1] 廖晓峰[1] 吴中福[1] 

机构地区：[1]重庆大学计算科学与工程学院,重庆400044 [2]重庆工业高等专科学校计算机系,重庆400050 [3]重庆教育学院计算机与现代教育技术系,重庆400067

出　　处：《重庆工业高等专科学校学报》2003年第1期4-8,共5页Journal of Chongqing Polytechnic College

摘　　要：传统的入侵检测方法在面对网络结构升级和未知攻击时 ,缺乏必要的扩展性和自适应能力 ,而基于机器学习的检测算法首先需要训练数据集进行训练 ,然后建立检测模型并通过测试数据集中入侵行为的检测结果来验证 ,此类方法由于获取类标识数据的困难性及其信息表达的局限性 ,降低了对未知攻击的检则能力。本文提出利用遗传聚类进行入侵检测算法IDUGC(IntrusionDetectionUsingGeneticClustering)。实验结果表明 ,此算法在未知入侵检测方面是可行的、有效的 。Traditional intrusion detection methods lack extensibility and adaptability in the face of upgraded network architectures and unknown attacks. Meanwhile, detection algorithms based on machine learning need train data sets to train firstly, then the detection module is set up and validated by the detection result to intrusions in test data sets. Because of difficulties in obtaining labeled data and limitations to the knowledge expression to labeled data sets, the ability to detect unknown attacks is finally degraded. In this paper, Intrusion Detection Using Genetic Clustering (IDUGC) is proposed. By means of simulated experiments, this algorithm is proved feasible, efficient and extensible for unknown intrusion detection.

关 键 词：入侵检 聚类 遗传算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]