DANDE: Deductive Anomaly Detection With Program Synthesis

Abstract : The core problem we are facing in this project is detecting qualitative anomalies. Qualitative anomalies are combinations of conditions or events that violate some standards of safety or normalcy. The essence of the current project is to coordinate multiple online data sources, including public records and Web-sites, using a theory of the relevant world knowledge, and deductive inference methods. The theory and methods are incorporated into Specware, the Kestrel software-development environment, Data sources include employment histories and qualifications, educational histories, Visa trails, organizations and affiliations, residence addresses, immigration records, radar tracks, security cameras, and flight-departure information. These sources, which are continually being updated, are linked to the Specware theory via a procedural attachment mechanism; this means that the inference mechanism can draw conclusions from the data sources just as if they were incorporated into the Specware theory.