基于多层分类器的恶意网页检测技术研究  

Research on Malicious Webpage Detection Technology Based on Multi-Layer Classifier

作　　者：张士坤 ZHANG Shi-kun(School of Computers,Guangdong University of Technology,Guangzhou 510006)

机构地区：[1]广东工业大学计算机学院,广州510006

出　　处：《现代计算机》2020年第18期64-68,共5页Modern Computer

摘　　要：互联网为人类社会带来便利的同时,也存在很多网络安全问题,例如钓鱼攻击、恶意软件、隐私泄露等。恶意网页在这些网络攻击中扮演重要的角色。针对网页不同层次特征,设计多层次的分类器检测模型。该模型采用CNN-GRU神经网络处理URL数据,采用随机森林算法处理网页特征数据,综合两个输出进行判定。实验证明,多层次检测模型识别准确率达到99%以上、且拥有更好的稳定性、更快的收敛性。While the Internet brings convenience to human society,there are also many network security issues,such as phishing attacks,malware,and privacy leaks.Malicious web pages play an important role in these cyber attacks.The article designs a multi-level classifier detection model for different levels of webpage features.This model uses CNN-GRU neural network to process URL data,random forest algorithm to process webpage characteristic data,and combines the two outputs to determine.Experiments prove that the recognition accuracy of the multi-level detection model is more than 99%,and it has better stability and faster convergence.

关 键 词：恶意网页 神经网络 随机森林 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]