(POS-47) Development of a Visualization Platform for Assessing Nuclear Proliferation Risks using Real-time Automated Open-source Data Collection and Deep Learning Technologies

Jae Min Lee - The Mirae Consulting Group
Soo Hwan Lee - The Mirae Consulting Group

The Mirae Consulting Group

Since 1970s, the assessment of nuclear proliferation risk has evolved significantly, strongly emphasizing technical factors such as proliferation resistance. Early methodologies including the Technological Opportunities to Increase the Resistance of Global Civilian Nuclear Power Systems (TOPS) employed an attribute-based approach that quantified physical properties, such as enrichment levels, radiation barriers, and safeguard measures. This approach was later expanded by frameworks such as the Generation IV International Forum (GIF) and the International Project on Innovative Nuclear Reactors and Fuel Cycles (INPRO), which incorporated additional technical dimensions. Subsequently, scenario-based approaches including the Markovian method have introduced pathway analysis to identify vulnerabilities, thereby enhancing nuclear nonproliferation assessment methodologies (NPAM). However, relying solely on technical factors is insufficient for a comprehensive evaluation of the nuclear proliferation risk. Political factors including state leadership intentions, international commitments, and geopolitical dynamics play a crucial role in determining nonproliferation credibility. To address this gap, integrated assessment methodologies that incorporate both technical and political dimensions have been developed. This paper proposes a platform for assessing nuclear proliferation risk levels that synthesizes these factors. Technical components are derived from established methodologies including TOPS, GIF, and INPRO, whereas political aspects are structured based on nonproliferation credibility frameworks. These factors were selected and refined using the Delphi method, and their relative importance was determined using an Analytic Hierarchy Process (AHP). To enable a real-time evaluation, a Java-based web crawler was developed to collect open-source information, initially focusing on North Korea. Several platforms including 38 North, Daily NK, and the Institute for Science and International Security (ISIS) were used as data sources. The collected data were processed and analyzed using natural language processing (NLP) and deep learning techniques to extract relevant insights. These processed datasets were then applied to a multi-criteria decision-making framework within the platform, ultimately generating visualized nuclear proliferation risk levels.