Real Time Malicious URL Detection on twitch using Machine Learning

Abstract

Live streaming systems like twitch enable broadcasters to live stream content to a huge audience. The Chatrooms are an important part of Twitch, allowing users to interact with the broadcaster, and among themselves freely. This means that these chatrooms are an ideal place for cybercrime attack and get sensitive information of the twitch user with the help of a malicious Uniform Resource Locators (URL). Harmful websites are responsible for a number of online cyber-attacks and scams today. The attacker tries to persuade unsuspecting twitch users to click on harmful URLs via Twitch Chatroom. Clicking on these URLs may result in the e-mail spam, the introduction of phishing scams, and the download of malicious software, adware, and ransomware, which may result in significant financial losses. As a result, it is very important to detect and prevent these threats effectively.The most common and fastest way to find malicious URLs is to help compare the URLs to the restricted list. However, blocking lists lack the ability to remove reproduced URLs. Unable to detect new bugs in the invisible URL of a strategy based on a completely restricted list, the machine-based reading strategy is integrated here to mitigate the above mentioned threats. The method uses lexical and semantic features extracted from a series of URLs, this feature extraction method is very secure and fast as it does not involve high computation as real-time protection is required.