This research proposes a systematic methodology for producing vulnerable aggregated analytics—summary statistics which, once released, enable partial reconstruction of the underlying dataset even when it has been anonymized—with the goal of raising awareness about the latent disclosure risks accompanying the publication of seemingly innocuous aggregates.
This talk will present the design of CoVault, a work-in-progress system for securely storing and querying data under a very strong threat model that doesn’t place trust in any one entity or authority, and includes the complete compromise of all CPUs of a specific manufacturer, as well as many common side channel attacks.