Research on multi-level malicious Web page Recognition based on topic Fusion and improvement of CNN

As one of the portals of the Internet, web application brings convenience to human society, but also has many problems, such as phishing attacks, malware downloads, privacy breaches, etc. In order to more accurately identify malicious web pages, reduce the risk of network attacks, this paper designs a multi-level detection model according to the different level characteristics of web pages. The model uses LDA-SECNN to learn the characteristics of web page content, uses random forest algorithm to learn the static characteristics of web page, and finally combines the two outputs to determine. Experiments show that the multi-level model can improve the accuracy of the model, and has good stability and convergence.