Studies and Analyses of Vulnerabilities in Aided Adversarial Decision Making
James Llinas, Colin Drury, Wayne F. Bialas, An-che Chen · 1998
Abstract : This report provides a preliminary examination of critical elements and processes involved in aided adversarial decision-making. The aid in the analysis (i.e., an automated decision aid) focuses upon a generic data fusion processor that estimates situation and threat states based on multisensor/multisource-based data assessments. In that frame of reference, this report provides a characterization of: (1) the information dependencies in data fusion processing, (2) the information dependencies in selected human-processing models, (3) the vulnerabilities of that information to Offensive 1W attack, and (4) the processes of decision making. It also examines prototypical cultural and technological differences among hypothetical adversaries in order to identify the a given adversary might place on specific information, and investigates the general nature of adversarial engagements in the context of a dynamic, two-sided, game-theoretic process. The report recommends that two critical areas of research be continued: (1) the study and modeling of information dependencies, vulnerabilities, and notions of value in automated decision making processes, and (2) the better understanding and modeling of patterns of human trust in automation. A case study framework is recommended, and, if more than one case study is defined, then an examination of the potential reuse of knowledge about critical information dependencies and values may be conducted.