Scalable machine learning framework for behavior-based access control

Abstract

Today's activities in cyber space are more connected than ever before, driven by the ability to dynamically interact and share information with a changing set of partners over a wide variety of networks. The success of approaches aimed at securing the infrastructure has changed the threat profile to point where the biggest threat to the US cyber infrastructure is posed by targeted cyber attacks. The Behavior-Based Access Control (BBAC) effort has been investigating means to increase resilience against these attacks. Using statistical machine learning, BBAC (a) analyzes behaviors of insiders pursuing targeted attacks and (b) assesses trustworthiness of information to support real-time decision making about information sharing. The scope of this paper is to describe the challenge of processing disparate cyber security information at scale, together with an architecture and work-in-progress prototype implementation for a cloud framework supporting a strategic combination of stream and batch processing.