USING BEHAVIOR ANALYSIS ALGORITHMS TO ANTICIPATE SECURITY THREATS BEFORE THEY IMPACT MISSION CRITICAL OPERATIONS 1 B. Banks, 1 G. Jackson, 1 J. Helly, 2 D. Chin, 1 T.J. Smith, 1 A. Schmidt, 3 P. Brewer, 1 R. Medd 1 Science Applications International Corporation (SAIC), Hawaii USA 2 University of Hawaii, Hawaii, USA 3 ObjectVideo, Virginia USA 4 D. Masters, 5 A. Burger, 6 W.K. Krebs 4 Department of Homeland Security 5 Alion Science and Technology 6 Office of Naval Research Virginia, USA Abstract The objective of this research is to identify, develop, adapt, prototype, integrate and demonstrate open access force protection and security technologies and processes. The goal is to provide more open public access to recreational and other non-restricted facilities on military bases and to improve the overall base safety and security utilizing advanced video and signal based surveillance. A testbed was created at the Pacific Missile Range Facility (PMRF), Kauai, Hawaii to demonstrate novel and innovative security solutions that serve these objectives. The testbed consists of (1) novel sensors (video cameras, radio frequency identification tags, and seismic, lidar, microwave, and infrared sensors), (2) a computer, data storage, and network infrastructure, and (3) behavior analysis software. The behavior analysis software identifies patterns of behavior and discriminates “normal” and “anomalous” behavior in order to anticipate and predict threats so that they can be interdicted before they impact mission critical operations or cause harm to people and infrastructure. 1. Introduction Following the attack on the USS Cole Guided Missile Destroyer (DDG-67), force protection became one of the United States Navy's primary concerns. Post-attack analysis recommended a need to collect, compile, coordinate, analyze, and disseminate threat information on a real-time basis to all relevant organizations. Implementing this recommendation with conventional force protection solutions would require significant manpower which conflicts with the Navy’s goal to reduce total ownership costs through reductions in personnel count. Manpower accounts for the largest percentage of costs across the lifecycle of a ship, aircraft, or system. These costs must be reduced if the United States Navy is to have the balanced resources to meet the war fighting challenges of the 21 st Century. Traditional video surveillance systems are susceptible to operators missing anomalous events, and they consume large amounts of manpower by requiring operators to monitor video displays for long periods of time. We proposed to investigate an alternative force protection solution that meets Navy’s manpower reduction requirement by relying on computational behavioral algorithms that analyze data from a sensor network and notify security personnel before a threatening event may occur on base. The software challenge is to develop a set of algorithms that will anticipate and predict threats and provide early warning to respond and interdict them. 2. Previous work There is significant past work on sensors, tracking, and data fusion [1] and on various behavior analysis methods [2–7]. A novel aspect of our research is the way our sensor based security systems are enhanced with advanced behavioral techniques. Initially our testbed will employ three behavior analysis methods: Automated Behavior Analysis (ABA), Agent Based Modeling (ABM), and Statistical Anomaly Detection (SAD). Sensor and Computer Systems. The trend in security solutions in 1990s was to move from analog to digital. Advances in computing power, video compression algorithms, digital storage, and data networking enabled this swift migration from analog to digital. The current trend is to move to intelligent and automated surveillance systems. The rapid expansion in the number of installed security sensors makes it no longer possible for all sensor outputs to be observed by security personnel. Furthermore there is a high demand for enhancing the advanced warning capability and the anticipation window. These requirements put more emphasis on utilization of advanced software algorithms for automated detection of threats, automated generation of alarms, and analysis of 978-1-4244-1696-7/07/$25.00 ©2007 IEEE.