Analysis of the Kim-Winkler Algorithm for Masking Microdata Files How Much Masking Is Necessary and Sufficient Conjectures for the Development of a Controllable Algorithm

By Moore, Richard A.

Book Id: WPLBN0000600487
Format Type: PDF eBook
File Size: 125,010 KB.
Reproduction Date: 2005

Title:	Analysis of the Kim-Winkler Algorithm for Masking Microdata Files How Much Masking Is Necessary and Sufficient Conjectures for the Development of a Controllable Algorithm
Author:	Moore, Richard A.
Volume:
Language:	English
Subject:	Government publications, Census., Census report
Collections:	U.S. Census Bureau Collection
Historic Publication Date:
Publisher:	U.S. Census Bureau Department


Citation APA MLA Chicago Moore, R. A. (n.d.). Analysis of the Kim-Winkler Algorithm for Masking Microdata Files How Much Masking Is Necessary and Sufficient Conjectures for the Development of a Controllable Algorithm. Retrieved from http://gutenberg.cc/

Description
Statistical Reference Document

Excerpt
Executive Summary: In 1995, the Department of Health and Human Services (HHS) contracted with the U.S. Bureau of the Census to produce a specially requested public use microdata file. This file supplemented information on the March 1991 Current Population Survey (CPS) public use file with Form 1040 information from each respondent?s 1990 Internal Revenue Service (IRS) tax return. Because the IRS could use their data to re-identify individuals on the CPS file, great care had to be taken to mask the HHS file. In doing so, Jay Kim and William Winkler (1995) developed a very effective two-stage procedure for masking public use microdata files. The initial step of this procedure adds randomly generated multivariate noise to the IRS data (Kim, 1986). Records are perturbed so that the means of the universe or any of its sub-domains are not biased. However, the noise is generated so that it increases the variances and covariances by a factor of (1 + c2) - where the masking agent specifies the value of c. For relatively ...

Table of Contents
Table of Contents Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 pages Chapters I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 II. Purpose of This Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 III. The Technique for the Measurement of the Amount of Masking Provided . . . . . . . . . . . 3 IV. Masking By the Addition of Random Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 V. The Test Deck and the Generation of Random Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Table 1. Statistics for the Fields Subjected to the Addition of Random Noise . . . . 5 Table 2. Statistics for the Random Noise For c = 0.100 . . . . . . . . . . . . . . . . . . . . 5 Table 3. Correlation Coefficients Before and After the Addition of Random Noise For c =0.1005 (R = 0.99) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 0 VI. The Masking Ability of Kim?s Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Table 4. Ability of Kim?s Addition of Random Noise to Mask Data ... . . . . . . . . . 7 VII. The Value Where c Loses Its Effectiveness, Mathematical Conjectures . . . . . . . . . . . . . 9 Table 5. Theoretical Value at Which c Loses Its Masking Power . . . . . . . . . . . . . 9 VIII. The Advantages and Disadvantages of Kim?s Masking Technique . . . . . . . . . . . . . . . . 11 IX. Winkler?s Matching Software, An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 X. The Advantages and Disadvantages of a Swap Procedure . . . . . . . . . . . . . . . . . . . . . . . 13 XI. Conjecture for a Re-Identification Tolerance Level, Historical Evidence . . . . . . . . . . . . 14 XII. Using Winkler?s Approach to Achieve Re-Identification Tolerance . . . . . . . . . . . . . . . . 16 XIII. Conjecture on an Optimal Masking Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 XIV. Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17