文件名称：”Better Than Nothing” Privacy with Bloom Filters：To What Extent？.pdf
文件大小：341KB
文件格式：PDF
更新时间：2022-10-29 07:11:45
Bitcoin
Bloom filters are probabilistic data structures which permit to conve- niently represent set membership. Their performance/memory efficiency makes them appealing in a huge variety of scenarios. Their probabilistic operation, along with the implicit data representation, yields some ambiguity on the actual data stored, which, in scenarios where cryptographic protection is unviable or unprac- tical, may be somewhat considered as a better than nothing privacy asset. Oddly enough, even if frequently mentioned, to the best of our knowledge the (soft) privacy properties of Bloom filters have never been explicitly quantified. This work aims to fill this gap. Starting from the adaptation of probabilistic anonymity metrics to the Bloom filter setting, we derive exact and (tightly) approximate for- mulae which permit to readily relate privacy properties with filter (and universe set) parameters. Using such relations, we quantitatively investigate the emerging privacy/utility trade-offs. We finally preliminary assess the advantages that a tai- lored insertion of a few extra (covert) bits achieves over the commonly employed strategy of increasing ambiguity via addition of random bits.