UNECE and FAO Dive Deep into Geospatial Data Confidentiality and Anonymization Techniques in Agriculture

From 26 to 28 September 2023, global statistical experts gathered in Wiesbaden for the UNECE Expert Meeting on Statistical Data Confidentiality. Coordinated by the Federal Statistical Office of Germany (Destatis) and RheinMain University of Applied Sciences, this meeting held significance in addressing challenges tied to the burgeoning demand for geospatial data derived from agricultural surveys.

In today's data-driven world, geospatial data, especially from agricultural domains, is witnessing escalating demand due to its varied applications spanning spatial statistical analysis, geographic cluster modeling, and the integration of farm data with remote sensed information. Yet, the dissemination of such data carries inherent risks, primarily centered around data privacy and potential disclosure.

One of the focal points of the conference was an abstract presented on innovative geo-masking techniques. The Demographic Health Survey (DHS) has been at the forefront with its geo-masking approach to ensure household location anonymity, a method that has since been recognized and employed by significant institutions like the World Bank.

Adding another layer to the discussion, the 50x2030 Initiative, steered by the United Nations Food and Agriculture Organization (FAO), aspires to fortify 50 low and lower-middle-income nations by 2030, equipping them with tools and strategies to produce, analyze, and use agricultural data in decision-making to bolster rural development and food security.

Within this extensive initiative, the FAO has been rigorously testing the applicability of DHS's geo-masking methods, specifically to agricultural holdings. Their findings underscore an underestimation of the actual risk, considering the nature of measures and the type of geospatial data used. The intricacies of designing a risk metric were elucidated, with emphasis on accounting for both spatial traits of agricultural holding locations and non-geospatial quasi-identifiers from survey microdata. As an alternative solution, FAO is investigating the dissemination of spatial variables extracted from anonymized coordinates, instead of the anonymized coordinates themselves. This method aims to enhance data utility while preserving privacy. However, a comprehensive analysis is necessary to assess the potential risks of grid cell re-identification associated with the dissemination of spatial variables through spatial signature.

The meeting was not just a platform for presentations but was designed to nurture collaboration, facilitate meaningful discussions, and, most importantly, drive the international coordination in the realm of statistical data confidentiality.