A Statistical Framework for Differential Privacy

• 成果类型: Article
• 署名作者: Wasserman, Larry; Zhou, Shuheng
• 署名单位: Carnegie Mellon University; Carnegie Mellon University; Swiss Federal Institutes of Technology Domain; ETH Zurich
• 刊物名称: JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
• ISSN/ISSBN: 0162-1459
• DOI: 10.1198/jasa.2009.tm08651
• 发表日期: 2010
• 页码: 375-389
• 关键词: disclosure

摘要

One goal of statistical privacy research construct a data release mechanism that protects individual privacy while preset ving information content An example is a random mechanism that takes an input database X and outputs a random database Z according to a distribution Q(n) (vertical bar X) Differential privacy is a particular privacy requirement developed by computer scientists in which Q (vertical bar X) IS required to be insensitive to changes in one data point in X This makes it difficult to inter front Z whether a given individual is in the original database X We consider differential privacy front a statistical perspective We consider several data-release mechanisms that satisfy the differential privacy requirement We show that it is useful to compare these schemes by computing the rate at convergence of distributions and densities constructed from the released data We study a general privacy method. called the exponential mechanism, introduced by McSheiry and Talwar (2007) We show dial the accuracy of this method is intimately linked to the rate at which the probability that the empirical distribution concentrates in a small ball around the true distribution