Chapter

Financial Cryptography and Data Security

Volume 5628 of the series Lecture Notes in Computer Science pp 32-50

Privacy-Preserving Information Markets for Computing Statistical Data

  • Aggelos KiayiasAffiliated withComputer Science and Engineering, University of Connecticut
  • , Bülent YenerAffiliated withComputer Science Department, RPI
  • , Moti YungAffiliated withGoogle Inc. and Computer Science, Columbia University

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Consider an “information market” where private and potentially sensitive data are collected, treated as commodity and processed into aggregated information with commercial value. Access and processing privileges of such data can be specified by enforceable “service contracts” and different contract rules can be associated with different data fields.

Clearly the sources of such data, which may include companies, organizations and individuals, must be protected against loss of privacy and confidentiality. However, mechanisms for ensuring privacy per data source or data field do not scale well due to state information that needs to be maintained. We propose a scalable approach to this problem which assures data sources that the information will only be revealed as an aggregate or as part of a large set (akin of k-anonymity constraints).

In particular, this work presents a model and protocols for implementing “privacy preserving data markets” in which privacy relies on the distribution of the processing servers and the compliance of some (a quorum) of them with the service contract. We then show how to compute statistical information important in financial and commercial information systems, while keeping individual values private (e.g., revealing only statistics that is performed on a large enough sample size). In detail, we present two novel efficient protocols for privacy-preserving S-moments computation (for S = 1,2,...) and for computing the Pearson correlation coefficients.