基于贝叶斯方法的单分类入侵检测技术  

One-class intrusion detection technique based on Bayesian approach

作　　者：肖仙谦[1] 朱俊平[1] 景旭[1] 马巧娥[1,2] 

机构地区：[1]西北农林科技大学信息工程学院,陕西杨凌712100 [2]杨凌职业技术学院信息工程学院,陕西杨凌712100

出　　处：《河北大学学报（自然科学版）》2014年第1期7-13,共7页Journal of Hebei University(Natural Science Edition)

基　　金："十二五"国家科技支撑计划项目(2013BAD15B02-3);中央高校基本科研业务费专项资金资助项目(QN2011036)

摘　　要：考虑到入侵检测问题中样本数据的不平衡性,可以将单分类模型应用到入侵检测中.通过应用贝叶斯方法,对单分类支持向量机模型进行改进,使之成为一种概率模型,从而更加符合入侵过程的随机特性;应用主成分分析技术对原始数据在各个方向上进行等方差处理,使之更加符合该模型对数据正态分布的前提假设.同时,在模型求解上采用分治的思想,对数据进行分组计算,实现对大数据的高效求解.在标准入侵检测数据集NSL-KDD上进行测试,实验结果达到87.46%的检测准确率,从而验证了该方法可以在入侵检测中得到有效利用.To consider the imbalance of the sample data in intrusion detection problem, one-class mod- els can be applied to intrusion detection. The one-class support vector machine（one-class SVM） was im- proved to be a probabilistic model by Bayesian approach, which makes it more conform to the random char- acteristics of intrusion process. The source data was preprocessed with the same variance in each orienta- tion by applying principle component analysis （PCA） technique,which makes it to be more suitable to the data normality hypothesis of the model. The idea of partition was applied in solving the model to calculate each data packet, so as to realize the efficient solution of large data. Test the model with the standard in- trusion detection dataset NSL-KDD, achieve the experimenl： result with 83.96 % detection accuracy, which validate the effectiveness of the techniques in intrusion detection application.

关 键 词：入侵检测 单分类支持向量机 贝叶斯估计 主成分分析 核函数 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]