Malicious Website Navigation Prevention Using CNNs and URL Vectors: A Study

In this paper, we have focused on the problem of malicious URLs. URL attacks have been on the rise in 2020, with most of the work being online based due to the pandemic there arises a greater scope of Phishing URLs etc. There have been existing systems but they are mostly paid, whereas with this project we aim to deploy a freemium add-on in a web browser, hosted on cloud with a real time dynamic classified URLs database so as to make the process more accessible and at the same time, less CPU and RAM consuming. The main highlights of our thesis have been that the accuracy measures of the two mains algorithms have been really close but there are discrepancies in the confusion matrix itself. Although these differences arise because of the time bindings and we would face such problems while deploying this project as an add-on service on a browser, a slow-fast multilayered system seems a better prospective plan to pursue in the future.