Data Recovery on Encrypted Databases with k-Nearest Neighbor Query Leakage

Explore a groundbreaking presentation on data recovery attacks targeting encrypted databases supporting k-nearest neighbor (k-NN) queries. Delve into the first-of-its-kind research that examines both unordered and ordered response scenarios, analyzing their vulnerabilities to exact and approximate reconstruction attacks. Understand the theoretical feasibility of exact reconstruction and the practical implications of approximate reconstruction techniques. Learn how these attacks can be applied to privacy-sensitive geolocation data using Hilbert curves for multidimensional spatial data mapping. Discover the potential risks and accuracy of these attacks through experiments conducted on real-world datasets, revealing relative errors ranging from 2.9% to 0.003% in reconstructing plaintext values.

Data Recovery on Encrypted Databases with k-Nearest Neighbor Query Leakage