Abstract

This research will propose a client side based phishing detection extension for Google Chrome browser that can assist in detecting and warning users about the websites they are currently browsing in realtime. This would be achieved by using random forest classifier a machine learning method, as based on past research done it concludes that it performs far better than other machine learning techniques in the field of detecting Phishing. Most common method of obtaining this is by providing the classification on a server and then performing a request to the server via the extension plugin to get the result. Unlike this method, this research will emphasize on running the classification on the browser instead. This is done because it has several advantages running on a client side as it provides a better privacy to the users since their browsing data and pattern is not compromised as it does not leave their machine. This will also make the detection plugin independent towards latency issues. This research mainly will focus on implementing machine learning in JavaScript for it to run on a browser as an extension since JavaScript does not have much library support towards Machine Learning and also to keep in mind of the users machines performance. This approach should be made with the intention of having it lite in order to achieve the capability to allow as much users as possible to use it. Random forest classifier for this project will be trained traditionally based on the phishing dataset 2 using Python scikit, and parameters of this model will then be exported in a JSON format to be used together with JavaScript.

Keywords: Phishing; Python; Machine learning; Java script