Contextual Reasoning for Threat Classification Refinement

Toyon proposes a two-pronged solution that minimizes risk to the customer while allowing exploration of novel techniques. Toyon will approach the problem from both a knowledge-based reasoning perspective and a Bayesian network perspective. The knowledge-based reasoning algorithms will cast the available information and structure as a knowledge graph and perform classification using reasoning from contextual entities. In this framework, each known training case object will act as an entity in the knowledge graph and will be connected both to other entities in its case as well as to other attributes of the entity, facilitating classification and discrimination. This approach will be compared to a lower-risk probabilistic statistical modeling of contextual data via Bayesian networks to represent conditional dependencies. Toyon will generate a scenario and accompanying datasets of features to test these algorithms, and will iterate on the reasoning framework given the results of the testing. Approved for Public Release | 21-MDA-11013 (19 Nov 21)