Click-Fraud Detection Using XG Boost and Convolutional Neural Networks

Abstract

With the growing usage of the internet, advertisers find this a convenient option to use this platform to advertise their apps or products. This gives many fraud publishers a chance to extract money from the advertisers in unethical ways. As most of these advertisers are displaying their content on the publisher’s page on a pay-per-click bases, it is very easy for the fraudsters to make fraudulent clicks and fool the advertisers into believing that their ads are actually being clicked by those many users. Hence the advertiser thinks that the app must have been downloaded   by most of the people who made the click if not all. So this is a kind of robbery that is going on in the cyber world. We, in this project are trying to compare the accuracy of prediction of two algorithms, XGBoost and Convolutional Neural Network to find out which one is best for this purpose. These algorithms will make a prediction if a particular click is a fraudulent one or will actually lead to a download of the app. This project deals with the detection of fraudulant clicks happening on the internet. Hence by doing this we are contributing our bit towards the cyber security domain.