Machine Learning Outlier Detection in Safetica's Data Loss Prevention System

Data loss prevention systems are becoming necessitiesin corporate computer system deployments. Nowadays, wheneverything is connected, and BYOD (Bring your own device)methodology is tolerated, even encouraged in many companies,network security administrators are obliged to keep with newesttechnologies to prevent threats to business resources. Threatsmight be parts of carefully planned corporate espionage, orsimple malware encrypting all resources available to it. No matterwhich threat, data have to be kept safe and each interaction withcritical business resources need to be monitored, authorized andlogged for future analysis.In this paper, we discuss state of the art methods used foroutlier detection, unsupervised learning and statistical analysis.