首页 | 本学科首页   官方微博 | 高级检索  
     

基于特征选择模型的钓鱼网站快速识别方法
引用本文:陈鹏,李勇志,余肖生.基于特征选择模型的钓鱼网站快速识别方法[J].计算机技术与发展,2021(4):40-45.
作者姓名:陈鹏  李勇志  余肖生
作者单位:三峡大学计算机与信息学院
基金项目:国家重点研究发展计划资助项目(2016YFC0802500)。
摘    要:目前在识别钓鱼网站的研究中,对识别速度有着越来越高的需求,因此提出了一种基于混合特征选择模型的钓鱼网站快速识别方法。混合特征选择模型包含初次特征选择、二次特征选择和分类三个主要部分,使用信息增益、卡方检验相结合以及基于随机森林的递归特征消除算法建立了混合特征选择模型,并在模型中使用分布函数与梯度,获取最佳截断阈值,得到最优数据集,从而提高钓鱼网站识别的效率。实验数据表明,使用该混合特征选择模型进行特征筛选后的数据集,维度降低了79.2%,在分类精确度几乎不损失的情况下,降低了32%的分类时间复杂度,有效地提高了分类效率。另外,使用UCI机器学习库中的大型钓鱼数据集对该模型进行评价,分类精确率虽然损失1.7%,但数据集维度降低了70%,分类时间复杂度降低了41.1%。

关 键 词:特征选择  信息增益  卡方检验  随机森林  递归特征消除

Method for Quickly Identifying Phishing Websites Based on Feature Selection Model
CHEN Peng,LI Yong-zhi,YU Xiao-sheng.Method for Quickly Identifying Phishing Websites Based on Feature Selection Model[J].Computer Technology and Development,2021(4):40-45.
Authors:CHEN Peng  LI Yong-zhi  YU Xiao-sheng
Affiliation:(School of Computer and Information,Three Gorges University,Yichang 443002,China)
Abstract:At present,in the research of identifying phishing websites,there is an increasing demand for recognition speed.Therefore,we propose a fast recognition method for phishing websites based on a mixed feature selection model.The mixed feature selection model consists of three main parts:primary feature selection,secondary feature selection and classification.A hybrid feature selection model is established by combining information gain,Chi-square test and recursive feature elimination algorithm based on random forest.The distribution function and gradient are used in the model to obtain the optimal cutoff threshold and the optimal data set,so as to improve the efficiency of phishing website recognition.Experimental data shows that the data set after feature selection using this mixed feature selection model has a 79.2%reduction in dimension,and reduces the classification time complexity by 32%with almost no loss of classification accuracy,effectively improving classification effectiveness.In addition,using the large-scale fishing data set in the UCI machine learning library to evaluate the model,although the classification accuracy rate is lost by 1.7%,the data set dimension is reduced by 70%,and the classification time complexity is reduced by 41.1%.
Keywords:feature selection  information gain  Chi-square test  random forest  recursive feature elimination
本文献已被 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号