NFDI₄Earth Call

We submitted an application for funding to the 2nd NFDI4Earth Call:

Privacy-compliant geosocial data stream transformation

Abstract Developments in recent years have shown that crowdsourced and geosocial data streams are important sources for understanding and predicting the effects of human activity on Earth, which is critical for solving problems in areas such as transportation, climate change, urban and landscape planning or disaster management. These data are generated on the basis of people’s participation and thus pose a particular challenge for data protection, especially when they are used for wider applications that are not known to the contributing user ahead of time. The goal is to advance existing approaches for privacy-preserving transformation of geosocial data streams, such as HyperLogLog (HLL), by using Locality Sensitive Hashing (LSH). LSH allows detailed modeling of spatial and semantic relationships without identifying individuals. Expected results are published methods for transforming crowdsourced and geosocial streaming data using LSH into privacy-preserving data structures, measures for determining the transformation properties (data volume, resolution, privacy budget) and generated benchmark data. For demonstration purposes, LSH is applied to geosocial data streams in the context of landscape and urban planning, e.g. to derive privacy-compliant indicators for ecosystem service monitoring. In the context of NFDI4Earth and research data management, availability and integration of these human-activity focused data sources plays an increasingly important role and the application of LSH for privacy-compliant transformation, storage, and use has the potential to foster a wide range of applications.

Figure 1. Illustration of the data lifecycle and the two cases of possible adversaries addressed by the privacy compliant transformation using LSH.

The total grant size was €71,410 for a one-year funding.