Malicious URL Prediction Using Machine Learning Techniques

Abstract

The usage of malicious website is a serious security threat faced by users while surfing data in internet. It is offensive and it belonging to criminal activities. Requirement of safeguard activities to help end-user is much needed. The need of understanding about protocol, uniform resource locator (URL) and other features of webpage are non-negligible. The purpose of this work is to findmalicious webpage from lexical and to resolve uncertainties faced by users. The study upon identifying features of websites is vulnerable and how malicious attack will occur is reported. To maximise the accuracy in prediction, machine learning technique is intruded. In recent years Phishing, botnet and malicious threats are more common in internet world and by disguising URL to trust it as non-offensive one.In concentrating with future concern and providing solution to real time problems faced by end-user the proposed work initiate two different algorithms namely decision tree and logistic regression. A total of 420000 webpages are taken as input data which included both affected and legitimate website. The time taken for prediction and accuracy is calculated in the form of testing data set. Thus logistic regression achieved higher efficiency with accuracy of 97.5% in an effective manner.