ZombieCam: Machine Learning common-sense anomaly detection for defense applications

ZombieCam software and technology, and the SERA anomaly detection algorithm, developed by Chiral Software in the course of our research for DARPA and Lockheed Martin, learns to detect anomalies in images and other data sets. Learning is based on common sense and observed context, not pre-defined rules. Video streams are understood for objects out of place, out of time, or otherwise unexpected. We plan to use this SBIR process to help us find applications for ZombieCam within the US Air Force and other government entities. The Air Force has needs for visual surveillance at a large scale. In addition, the same SERA anomaly detection algorithm could be adapted to find attacks or irregularities in network traffic, file storage access, and other areas. Detection of anomalies in other domains could be used to prevent network attacks and attempts to exfiltrate data. Data exfiltration cannot successfully be blocked with conventional approaches of user permissions and audit logs, but could be detected and stopped with the SERA algorithm. As a small, independent business, the SBIR special program, with the MD5 Accelerator, is a perfect fit to help us bring our innovative technology to a wide range of government users.