As we are not cybersecurity experts, we had a lot of difficulties in the beginning, but by using our general data science knowledge and a bit of research, we managed to process the provided data in a way we think could be valuable.
We tried different methods of clustering and outlier/anomaly detection, including a home-made one that tries to build on the scientific paper about formal concept analysis, provided by the university.
Three outlier-detection methods have been implemented, all of which are meant to guide forensic analysts towards data points that could be of high interest. Due to a lack of control data the parameters of our methods could not be fine-tuned. In case of real world use we expect the optimization of parameters to significantly improve the results.
Log in or sign up for Devpost to join the conversation.