TY - GEN
T1 - Asymptotically optimal and private statistical estimation
AU - Smith, Adam
N1 - Funding Information:
Supported in part by US National Science Foundation (NSF) TF award #0747294 and NSF CAREER/PECASE award #0729171.
PY - 2009
Y1 - 2009
N2 - Differential privacy is a definition of "privacy" for statistical databases. The definition is simple, yet it implies strong semantics even in the presence of an adversary with arbitrary auxiliary information about the database. In this talk, we discuss recent work on measuring the utility of differentially private analyses via the traditional yardsticks of statistical inference. Specifically, we discuss two differentially private estimators that, given i.i.d. samples from a probability distribution, converge to the correct answer at the same rate as the optimal nonprivate estimator.
AB - Differential privacy is a definition of "privacy" for statistical databases. The definition is simple, yet it implies strong semantics even in the presence of an adversary with arbitrary auxiliary information about the database. In this talk, we discuss recent work on measuring the utility of differentially private analyses via the traditional yardsticks of statistical inference. Specifically, we discuss two differentially private estimators that, given i.i.d. samples from a probability distribution, converge to the correct answer at the same rate as the optimal nonprivate estimator.
UR - http://www.scopus.com/inward/record.url?scp=71549153186&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=71549153186&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-10433-6_4
DO - 10.1007/978-3-642-10433-6_4
M3 - Conference contribution
AN - SCOPUS:71549153186
SN - 3642104320
SN - 9783642104329
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 53
EP - 57
BT - Cryptology and Network Security - 8th International Conference, CANS 2009, Proceedings
T2 - 8th International Conference on Cryptology and Network Security, CANS 2009
Y2 - 12 December 2009 through 14 December 2009
ER -