摘要
针对网络攻击检测准确率较低的问题,提出基于人工神经网络和遗传算法的混合网络攻击检测算法。将多目标遗传算法和多项式逻辑回归模型组合成封装特征选择算法,利用多项式回归模型对多分类数据的高效学习能力以及多目标遗传算法的全局优化能力,提取数据的最优特征子集;将降维后的特征集送入感知机训练,利用重引力搜索算法搜索神经网络的参数。基于不同的网络数据集完成实验,实验结果表明,该算法有效降低了特征维度,实现了较好的检测性能。
Aiming at the problem of low network attack detection accuracy,a hybrid network attack detection algorithm based on artificial neural networks and genetic algorithm was proposed.The multiple objective genetic algorithm and the multinomial logistic regression model were combined as a wrapper feature selection algorithm,the efficient learning capacity of the multinomial logistic regression model and global optimization capacity of the multiple objective genetic algorithm were utilized,and the optimal feature subset was achieved.The dimension reduced feature sets were delivered into perceptron to train,the gravitational search algorithm was adopted to search the parameters of neural networks.Experiments were carried out on different datasets,results show that the proposed algorithm reduces feature dimension effectively,and it also achieves good detection performance.
作者
罗予东
陆璐
LUO Yu-dong;LU Lu(School of Computer Technology,Jiaying University,Meizhou 514015,China;School of Computer Science and Engineering,South China University of Technology,Guangzhou 510641,China)
出处
《计算机工程与设计》
北大核心
2021年第9期2446-2454,共9页
Computer Engineering and Design
基金
国家自然科学基金青年基金项目(61402178)。
关键词
网络安全
入侵检测
特征选择
多项式逻辑回归
不平衡数据分类
多层感知机
networks security
intrusion detection
feature selection
multinomial logistic regression
imbalanced data classification
multi-layer perceptron
