Using Machine Learning for User Access Management

In any enterprise, one of the most prevalent security risks revolves around who has access to which resources. Whether data is being stored in a cloud solution or on-premises, there is a large challenge in knowing how to provide the correct privileges to associates. By using machine learning and clustering algorithms like the Louvain Method, we can group similar users in the Capital One network and create two valuable features: (1) automated onboarding and (2) automated “rogue access” detection. With the utilization of machine learning, we have allowed Capital One to become a more well-managed company, and have reduced a major cybersecurity threat. This talk will be a deep dive into the model, data engineering and productionization of the web application interface.

I am a machine learning engineer working for Capital One. I graduated from MIT in 2015 with a bachelor's degree in Mechanical Engineering and I am currently pursuing a master's degree from Georgia Tech in Artificial Intelligence and Machine Learning.