How can data that saves lives in crisis situations be shared without putting people at risk? Researchers at the Centrum Wiskunde & Informatica (CWI), located at Amsterdam Science Park, have developed an innovative solution for the United Nations (UN) to filter sensitive information from its global crisis data platform HDX (Humanitarian Data Exchange).
Together with master’s student Liang Telkamp, researcher Madelon Hulsebos created two mechanisms to automatically detect “contextually sensitive” data in the thousands of datasets shared on HDX. This includes information that—depending on time, place, or role—can potentially harm people in vulnerable positions. Think of hospital coordinates in war zones or personal data of displaced persons.
The researchers used large language models (LLMs), including GPT-4 and open-source variants, to automatically apply UN guidelines to diverse data sources. This AI reads the rules, interprets the context, and determines which data points might be potentially harmful. The results are impressive: the system detects nearly 94% of sensitive personal data—significantly outperforming the current standard tool (Google DLP), which identifies around 63%.
The UN has decided to integrate this new method into its HDX platform. As a result, the work of Hulsebos and Telkamp directly contributes to better protection of individuals in crisis situations. Hulsebos will present the research in October at a UN meeting in Barcelona, where international organizations gather to improve data management in the humanitarian sector.
Read the full article on the CWI-website
Looking for partners to collaborate. Or looking for a certain expertise? Or would you like to locate your business in the Amsterdam Science Park? Drop us a line and we help you to find a perfect match.
Subscribe to our LinkedIn-newsletter and we will keep you updated on all that our park has to offer. It will arrive to your inbox five times a year and you can unsubscribe easily at any time.
Subscribe