Information Theory and Privacy in Data Banks,
Abstract
The problem of providing privacy transformations for data banks and retrieval systems falls into two modern disciplines: information theory and computer science. In the paper the concern is primarily with the former. Here it is shown that measures of privacy can be drawn from information theory and, in particular, from rate distortion theory. It is demonstrated that a rate distortion function or the minimum mutual information about a record, conveyed by a distorted version of the record, is a natural measure of privacy. Shannon's conditions for perfect secrecy are met if the average mutual information about an original record R, given a distorted version E, is zero. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1973
- Accession Number
- AD0762577
Entities
People
- I. S. Reed
Organizations
- RAND Corporation