semSCI - Semantic Application to Detect and Resolve Suspicious and Conflicting Information

ABSTRACT: In the current security environment, violent extremist organizations are comprised of global networks of loosely connected cells marked by centralized decision making but decentralized execution of operations, where individuals are increasingly adept at leveraging various forms of communication, transaction mechanisms, and travel patterns in support of malicious agendas. Within these multiple layers of information, intelligence analysts require a capability to detect and resolve conflicting, inconsistent, suspicious, and deceptive data, reducing the uncertainty in analysis associated with misinformation. In response, we are proposing to develop semSCI, a Semantic Application to Detect and Resolve Suspicious and Conflicting Information that enables analysts to combine diverse sources of structured and semi-structured information within a common schema to automatically tag entities and relationships, including metadata about provenance such as timeliness and reliability. semSCI will represent the asserted facts in the structured and semi-structured information using a semantic annotation formalism to create a knowledge graph data model. Leveraging this knowledge graph, semSCI can infer not only spatial, temporal, and naming conflicts but any inconsistency indicating suspicious and deceptive information involving the logical expressions of subject and property values in the multi-dimensional semantic space with the use of stream entropy algorithms. BENEFIT: This project will result in the development of software products for the data management for intelligence market, supporting the integration of semi-structured and structured data from a variety of sources to include highly technical data formats for the purposes of identifying suspicious, conflicting, deceptive, and inconsistent information. Given the difficult budget climate, DoD is leaning toward multi-purpose technologies that fuse various collection disciplines and standardize reporting. semSCI is directly in line with this focus, as our DL based solution can fuse various data formats by incorporating the underlying semantics of the data into the ontology. In alignment with DoD strategy, semSCI will focus on special operations, as well as intelligence, surveillance and reconnaissance equipment, unmanned systems, space systems and cyberspace tools. There is considerable commercial opportunity in applying this technology to the homeland security context as well, whereby users would be filtering incoming sensor feeds such as social media artifacts, data from national and local government organizations, and weather information for building a common operating picture to respond to natural disasters and unconventional threats. Detecting conflicting, suspicious, deceptive and inconsistent data within these multiple layers, especially within social media, could be critical for first responders and policymakers in responding to a crisis. In the enterprise segment, we intend to commercialize the proposed technology by developing a cyber intelligence service, whereby the solution would fuse various types of cyber data to a common ontology and detect inconsistencies, and conflicting, suspicious, and deceptive data.