SBIR-STTR Award

Military installations operating in forward locations face unique challenges when attempting to automate threat detection in security monitoring. Due to complex interactions with allied and adversarial groups and the temporary nature of many such bases, development of a general framework that can identify emerging security threats in these dynamically changing environments poses an array of difficult technical obstacles. DECISIVE ANALYTICS Corporation (DAC) proposes the development of the Discovery of Anomalous Activity in Real Time (DAART) system. The DAART system is unique in its approach to threat monitoring in that it requires no supervised data to be trained for use at a new facility. Interpretations of incoming sensor information and detection of anomalous activity will be learned through strictly unsupervised machine learning methods at each base where the system is deployed. This allows the system to be highly adaptable and easily customized to the individual environment of each base, and lowers the requirements to deploy the system in new locations. Classification of threats will be learned using a semi-supervised machine learning technique in collaboration with operators, which will allow an extremely high level of control over threat detection and classification, while also minimizing the false alarm rate of the system.

Military installations operating in forward locations face unique challenges when attempting to automate threat detection in security monitoring. Due to complex interactions with allied and adversarial groups and the temporary nature of many such bases, development of a general framework that can identify emerging security threats in these dynamically changing environments poses an array of difficult technical obstacles. DECISIVE ANALYTICS Corporation (DAC) proposes the development of the Discovery of Anomalous Activity in Real Time (DAART) system. The DAART system is unique in its approach to threat monitoring in that it combines the power of unsupervised learning techniques for detecting unknown and unanticipated threats, with supervised learning techniques that excel at detecting high-value known threats. This combination of approaches provides an unprecedented facility security system that is flexible, robust, and highly adaptable to any sensor type. The system is easily customized to the individual environment of each base, making deployment of the system to new facilities easy. DAART detects emerging threats in real time, intelligently combining information from disparate sensors to help operators quickly understand the nature of each detected threat, enabling rapid responses to emerging threats.