https://www.crowdstrike.com/blog/how-we-trained-overfit-models-to-identify-malicious-activity/