Behavior Signatures

The objective of this research project is to develop a new class of multi-attribute behavior signatures to enable the anticipation of enemy activities. Asymmetric warfare and operations against transnational terrorist groups would significantly benefit from a new class of multi-attribute behavior signatures for predicting adversary intent and their likely courses of actions. These behavior signatures could potentially draw upon multiple streams of intelligence data, possibly over long temporal durations, to provide direct and indirect indicators of activities of interest. Conceptually, behavior signatures can be considered to be schemas whose attributes delineate a set of characteristics, or patterns of activity, that define the behavior of potential threat entities (consisting of individuals, groups, organizations, societies, and/or nations/states). The activation or realization of a behavior signature for a particular threat entity can be all or nothing or partial, where some, but not all, of the attributes are matched. Research is needed to define selected initial signature libraries, to explore the type of architecture required to use behavior signatures as a computational system (such as probabilistic neural networks, multinomial logistic regression, etc.), and to identify which Air Force systems would benefit most from behavior signatures technology (such as counter intelligence from influence operations).