Bot or Human: Identifying Bot­Generated Clicks Using Machine Learning

Abstract

“I’m not a robot” is a common CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) often shown today upon entering websites. The purpose behind the challenge is to distinguish hu mans from robots, or bots. However, the user experience becomes somewhat intrusive and is not always viable for many websites. This project explores, in collaboration with Prisjakt, how to retroactively identify clicks generated by bots, using historical data and various machine learning models. The models are trained and evaluated on the historical data in an effort to be able to classify future clicks automatically. The result of the project is an implementation of two models, a neural network and a gradient boosting model, as well as an application programming interface (API) to use the models with. The models show very promising results and suggest that an automated system, which identifies clicks generated by bots, is possible.