Abstract: To improve network intrusion detection rate, this paper proposes a network intrusion detection algorithm based on joint optimization algorithm for feature selection and classifier. The algorithm uses the relationship between the feature selection and classifier parameters. The feature and classifier parameters are taken as genetic algorithm individuals while the network intrusion detection rate as the evaluation function of genetic algorithm, then the optimization parameters are obtained by selection, crossover and mutation of genetic algorithm, and the joint optimization algorithm is tested by KDD 1999 data. The experimental results show that the proposed algorithm has overcome the degression of detection ability resulting from unmatched of the features and the classifier parameters, and has improved the network intrusion detection rate compared with other algorithms; it provides a new way for network intrusion detection.

Key words: network intrusion, feature selection, Support Vector Machine（SVM）, Genetic Algorithm（GA）

摘要： 为了提高网络入侵检测正确率，利用特征选择和检测分类器参数间的相互联系，提出一种特征和分类器联合优化的网络入侵检测算法。联合优化方法将网络状态特征和分类器参数作为遗传算法的个体，网络入侵检测正确率作为个体适应度函数，通过选择、交叉和变异等遗传操作获得最优特征和分类器参数，利用KDD 1999数据集对联合优化算法进行验证性测试。实验结果表明，相对于其他入侵检测算法，联合优化算法既解决了特征与分类器不匹配带来的入检测检测能力下降，又提高了网络入侵检测正确率和效率，为网络入侵检测提供了一种新的研究思路。

关键词: 网络入侵, 特征选择, 支持向量机, 遗传算法