Multiple Threat Tracking and Analysis (MULT2A)

Creating Situational Awareness (SA) involves the collection, fusion, analysis, and dissemination of data and information. This data must be developed into discoverable information and delivered to the appropriate decision makers within their decision timeline. It currently takes considerable time and skill for analysts to characterize SA and emerging threat patterns. A primary underlying cause is that significant relationships among entities and events are not easily discernible. Moreover, sequences of events and associations are formed by hypotheses based on individual judgment, given limited a priori knowledge since some behaviors and events have few or no direct observations. Additionally, the overwhelming amount of multi-INT data from diverse sources contain information that is often missing, wrongly associated, or obscured by “noise”. The result is obscuration of relevant threat patterns. Changes in behavior patterns over time may be stealthy enough to avoid immediate human detection. These indicators, represented as spatial and temporal sequences, can be interpreted as indicators of an impending adversary action to disrupt or degrade a US asset. The proposed technology defines a novel machine learning and data fusion-based architecture that enables analysts to autonomously identify and analyze multiple simultaneous perceived threats based on learned past behavior coupled with the fusion of streaming multi-INT, multi-source data. The technology provides analysts the ability to draw conclusions about adversary activities and intent, to assess the likelihood and timing of potential threats to US assets and warfighters, and to make recommendations for sensor tasking and/or appropriate actions within available response time constraints.