Alexandra Mendes, INESC TEC researcher, is one of the winners of the Atlantic Security Award, promoted by FLAD; Alexandra’s work aims to use a large language model – trained with data retrieved from the dark web – in decision-making processes, in the design of Defense policies and strategies, and in the application of the law in the security of the Atlantic region.
In the last months of 2023, Alexandra Mendes – at the time, a Visiting Researcher at Carnegie Mellon University (U.S.A.) – was at a weekly meeting of the research group led by Prof. Nicolas Christin, when she came across a research project in which price fluctuations in drug markets were analysed using data from the dark web. Many criminal actions take place in this “dark” section of the Internet, featuring intentionally hidden encrypted networks.
Some months later, the researcher came across the Atlantic Security Award – promoted by the Luso-American Development Foundation (FLAD), the Atlantic Centre, and the National Defense Institute (IDN) – and decided to introduce into this solution an LLM, i.e., an AI system capable of understanding and generating human language to process large amounts of data.
According to Alexandra Mendes, this was “an excellent opportunity to explore how LLMs could improve the analysis of the vast data available on the dark web”, together with prompt engineering techniques, which are based on the structuring of instructions that can be interpreted and perceived by generative AI models. These were the first steps towards a proposal – which would eventually win the award.
This research will produce a prototype that uses a trained and/or fine-tuned LLM to handle a comprehensive set of dark web data and, consequently, contribute to security in the Atlantic. At the same time, this data should support the formulation of Defence strategies and policies, and the planning of operations against cybercrime, illegal trade and other illicit practices that occur in the Atlantic, and that end up emerging in this uncontrolled online section.
“To effectively tackle criminal activities online, and contribute to security and defense policies, it is essential to assess the ongoing threats on the dark web”, explained Alexandra Mendes. “This is not a simple effort, due to the difficulties in collecting and extracting meaning from the various transactions that occur”.
According to Alexandra Mendes, this award is an “important milestone” in her career, recognising her professional path and the value of the research proposal. The Atlantic Security Award was created in 2021, awarding each selected researcher €15K to develop their research project over eight months.
The researcher mentioned in this news piece is associated with INESC TEC and UP-FEUP.