Understanding Anomalous Network Behavior with AI and ML

Maintaining software infrastructure for business support systems can be a manually intensive and unsustainable process, that would greatly benefit from AI-based automation. Such systems can comprise networks of hundreds of servers, distinct APIs, and tracked metrics. An overwhelming and intractable amount of data for a human support engineer to parse. To augment their capabilities, machine-learning-based anomaly detection techniques and AI-enhanced root cause analysis can be used to improve situational understanding and speed up incident resolution times. In this talk we review our work in developing data-science-based tools for understanding and predicting anomalous network behavior in software infrastructure.