A fuzzy logic decision support system for open source information analysis in a non-proliferation framework

Full text: In the last decades events in the international scene forced a radical enhancement in the mission of the IAEA: from verifying the non-diversion of safeguarded material to giving assurance about the absence of undeclared material, in compliance with States' safeguards undertakings. For that purpose, a new legal and technical tool was open to signature: the Additional Protocol, which requires States to provide the IAEA with a wider range of information about materials, equipment, know how and sites directly or indirectly related to the nuclear fuel cycle. Subsequently, the declarations provided by the States are confronted with inspections' results and other information sources, including open sources, in order to assess both their correctness and completeness. Hence, the information analysis process, and especially the analysis of open source information has become a crucial aspect of the State Evaluation Process. A decision support system (dss) is proposed as a tool that supports the analyst in the acquisition of relevant information from open sources and that performs a synthesis of the information acquired. Based on the most comprehensive set of proliferation indicators available, the IAEA's Physical Model (PM), the dss breaks up in two components: an acquisition and analysis module where information is drawn from open sources and an aggregation and output module. In the acquisition and analysis module, a semantic search tool (topic tree) is developed based on the Physical Model, that supports and narrows the results of the search of information among open sources, typically the internet. Subsequently, according to what is stated in the documents found, experts assign values to the related Physical Model indicators. In order to adapt to the specific nature of open source information, imprecise, uncertain and often lacking, indicators are assigned linguistic values, which are stored in the indicators' database. The aim of the synthesis and output module is to combine the large quantity of (linguistic) information stored in indicators into a few synthesis (numerical) values. Since fuzzy set theory was studied as a logical system for computing with words, it appeared most natural to develop the synthesis methodology in the framework of fuzzy sets theory. The methodology is based on three principles: transparency, multiplicity of sources and traceability of information. It follows a two-phase 'human-like' aggregation process taking into consideration the reliability of the sources providing the information and the relevance of the indicator in the proliferation path. Each phase makes use of specific aggregation tools: combination operators for mathematical aggregation are employed when performing aggregation according to reliability, whereas a fuzzy inference system, that is to say knowledge based tool is employed when performing aggregation according to relevance. Finally, an output tool is outlined based on a graphical representation of the Physical Model that allows an immediate visualisation of the overall nuclear fuel cycle where the technical capabilities acquired are highlighted. A decision support system has been developed based an original methodology as a contribution to enhance the effectiveness of the information analysis process, and, to some extent, the efficiency of the State evaluation process which is today at the centre of safeguards systems. By providing a synthesis of conspicuous amounts of information, the decision support system considerably reduces analytical work at the lower level, while experts' evaluation capacity at a higher level is not questioned. Further enhancements are foreseen to consolidate the present proposal. One issue to be faced is the continuous update of the incoming information, which is expected to be evolving at a very rapid pace. Moreover, including a quantitative dimension would allow performing material balance consistency checks. Finally, the set of indicators taken into account here is strictly technical, but it could be complemented with non-technical indic ators, such as economic or political ones, in order to enlarge decision-makers' perspectives. (author)