PRECOG: Pattern-Based Risk Evaluation through Computation of Outlier Groups

In the current security environment, where much of the conversation takes place in online communities and digital footprints are left from a variety of transactional data sources, piecing together the disparate connections that form a pattern-of-life is a challenging problem. What is required to address this problem is to provide unsupervised learning approaches that can identify important shifts in patterns-of-life across disparate datasets without human input to generate alerts to analysts of individuals who may have an intent to harm. In response to this challenge, Aptima, Inc. and our partners, Smart Information Flow Technologies (SIFT) and information operations specialist Lt Col Janelle Ward, USAF (Retired), propose to develop PRECOG, Pattern-Based Risk Evaluation through Computation of Outlier Groups. PRECOG will aid intelligence and information operations analysts in detecting outliers in dark patterns of life to predict individuals and groups at risk for harmful action across multimodal datasets through the innovative integration of algorithmic approaches. This capability advances the state of the art by integrating a number of approaches into a multidimensional machine learning and reasoning capability to include automated intent-to-act adversely linguistic features, automated temporal activity graph generation for pattern-of-life characterization, and unsupervised learning approaches for detecting outliers in multimodal datasets.