Robust Representation: Real-time Characterization of Streaming Data

The volume and velocity of digital data collected and used for analysis is growing at an increasing rate and will soon exceed our ability to store and analyze it in a timely fashion.Additionally, much of these data represent changes in behaviors of the underlying networks and entities that they measure such that the rate of development of new analytics to keep up with events and trends, such as changing cyber attack vectors, is severely limited by available tools.We propose to reduce the need for data storage to only those data needed for understanding by developing a system that tunes processing and storage needs to human cognitive capabilities.In this Phase II SBIR we will use the Robust Representation approach to develop statistics that can take data incrementally and usefully detect events, trends, and baseline behaviors.We anticipate that this system will be a commercially useful tool for a large variety of real-time analytic needs.In particular, we will target commercial cyber network defense as an initial customer base as it has a critical need for real-time analytics to detect and respond to cyber attacks.