DATA FUSION & NEURAL NETWORKS, LLC — Department of Defense SBIR Phase I: AF162-004

Description

DF&NN and MarkLogic will enable fast search and retrieval of big data. Detection and characterization processes both for real-time and after-the-fact analysis will maintain themselves based upon user goals and thus will not require substantial expert interaction to create and maintain rule sets. The resulting new Enterprise Ground Satellite Command and Control (E-GSCC) system will learn normal behaviors in real satellite telemetry off-line and then in real-time provide historical pattern matching and unknown abnormal pattern discovery across missions and threat locations. The E-GSCC detection and reporting processes will decide when to retrain, what data to retrain on, what to test on, and when to promote to on-line operations. Operators will monitor processes, set thresholds in real time for detection and reporting of events with supporting correlations to historical events, likely attribution, and response recommendations. We have experience in delivering these types of capabilities at TRL7 which reduces risk. We will leverage these capabilities to index, export, and search large volumes of archived data, across streams of telemetry and mission data from multiple satellites to demonstrate high performance searching for valid disparate balanced training set selection, off-line data-pattern learning, and on-line abnormal measurand correlation detection, event cause characterization, and response recommendations.