Proactive intelligence for nuclear nonproliferation

The work described in this paper leverages predictive models for proliferation detection in order to assess the complementary questions of capability and intent as they relate to the potential for nuclear weapon development. The ability to proactively assess the likelihood of a state to engage in nuclear power acquisition and development for non-peaceful purposes is one of the greatest challenges for analysts and policy makers working on proliferation detection and deterrence. Of further difficulty is determining whether a state is at risk to provide indirect support for proliferation via the relationship between industrial input/output and the legal framework of trade. In general, it is possible to gather evidence about precursor activities to the achieved nuclear potential of a state that function as indicators of the state's intent to acquire and develop capabilities to support nuclear weapons. Reasoning with these indicators to predict intent and capability to proliferate is of utmost importance to facilitate nuclear safeguards, e.g. through proactive implementation of countermeasures. Such a predictive reasoning task is difficult to perform without computational aid. While the need for a proactive and multi-perspective approach to proliferation detection is widely recognized, there is a lamentable lack of computational tools applied directly to the task. Applications of predictive modeling to the domain of nuclear nonproliferation are limited to physical/chemical properties of nuclear materials, such as nuclear weapons simulations and stockpile stewardship. Our aim is to address this gap by leveraging methods and data from different mission areas in support of proliferation detection and prevention in innovative ways. More specifically, the approach we have developed combines methods in information analysis and probabilistic evidentiary reasoning with expert knowledge from discipline areas germane to the proliferation detection, and evidence extracted from relevant data sources to assess alternative hypotheses about specific proliferation detection problems.