Inferential Disclosure Risk in Computer Databases

Principal Investigators: George T. Duncan and Ramayya Krishnan

The Heinz School of Public Policy and Management, Carnegie Mellon University

1. BRIEF SUMMARY OF PROGRESS ON IRI-9312143

There are several related research projects being undertaken with the NSF funding. They are summarized below. Each project summary has a brief abstract, and lists the faculty and students working on the project. It also lists papers and conference presentations related to the projects.

A. STATISTICAL DISCLOSURE LIMITATION IN COMPUTER DATABASES

We are currently investigating disclosure detection and protection of multivariate categorical databases. These find wide application in such areas as health care management. We have designed and developed models and algorithmic methods. These are reported in a technical report titled "Disclosure Detection and Protection for Multivariate Categorical Databases".

Faculty working on the project
George Duncan (PI)
Ramayya Krishnan (Co-PI)
Sumitra Mukherjee, City University of New York
Stephen Roehrig, Carnegie Mellon University

Graduate Student working on the project
Sumit Dutta-Chowdhury

Conference Presentations and Papers
Duncan, George T., Krishnan, Ramayya, and Mukherjee, Sumitra. Inference channel detection in multilevel relational databases: a graph-based approach. To appear in the Journal of Organizational Computing.

Duncan, George T. and Mukherjee, Sumitra. Confidentiality within computer databases. To appear in Statistica Applicata.

Duncan, George T., Mukherjee, Sumitra, "Data Access and Confidentiality in Statistical Databases", to be presented at the meeting of the International Society for Decision Support Systems (ISDSS), Hong Kong, June 1995

Duncan, George T., S. Dutta-Chowdhury, Krishnan, Ramayya, Roehrig, Stephen, "Disclosure Detection and Protection for Multivariate Categorical Databases", Technical Report, The Heinz School Carnegie Mellon University, 1995. This paper will be presented at the Institute of Management Science (now merged with the Operations Research Society of America to form INFORMS) International Meeting in Singapore, June 1995.

Duncan, George T. and Mukherjee, Sumitra, Access and confidentiality in statistical databases: optimal disclosure limitation through additive noise. Submitted to Journal of the American Statistical Association.

Duncan, George T. (1994) Invited presentation, Perspectives on "Private Lives and Public Policies," Senior Executive Meeting, U.S. Bureau of the Census, Suitland, MD, April 20.

George T. Duncan, (1994) Keynote address, Private lives and public policies: Two pillars of the global information highway, Second Annual Research Symposium of the Council for Marketing and Opinion Research, New York City, June 30.

George T. Duncan, (1994) Invited roundtable discussion leader, Ramifications of "Private Lives and Public Policies," Annual Meeting of the American Statistical Association, Toronto, August 15-18.

George T. Duncan, (1994) Invited participant, Workshop on Ethical, Legal and Technological Aspects of Computer and Network Use and Abuse, American Association for the Advancement of Service/American Bar Association, Queenstown, MD, October 7-9.

George T. Duncan, (1994) Invited talk, Incentives to respondents: Towards a productive and ethical policy, Institute for Social Research, University of Michigan, October 18.

George T. Duncan, (1995) Special Invited Address, (1995) Is there a database in our future? Twentieth International Symposium on Computer Science, El Instituto Tecnológico y de Estudios Superiores de Monterrey (ITESM), Monterrey, Mexico, February 23-25.

B. WORLD WIDE WEB-BASED MARKETS FOR COMPUTATIONAL SERVICES

This project is designing and developing electronic web-based markets for computational decision support services. Databases which capture information about transactions taking place in the market are natural by products of these types of markets and need to be adminstered taking into account data access and confidentiality issues.

Faculty Working on the Project
Ramayya Krishnan (PI)
Hemant Bhargava, Naval Postgraduate School, Monterey
Rudolf Muller, Humboldt University, Germany

Graduate Student Working on the Project
David Kaplan

Conference Presentations and Papers
Gunther, Oliver, Koerstein, Rolf, Muller, Rudolf, Schmidt, Peter, Krishnan, Ramayya, The MMM project: Access to Algorithms over the WWW, Online paper to be published in the Proceedings of the Third International World Wide Web Conference, Darmstadt, Germany, 1995. Available at http://coltrane.wiwi.hu-berlin.de/mmm_poster.html

Krishnan, Ramayya, Muller, Rudolf, Scmidt, Peter, "On accessing computable information over the WWW: The MMM project", Technical Report, The Heinz School, Carnegie Mellon University, 1995; to be presented at the ISDSS conference in Hong Kong, June 1995.

Bhargava, Hemant, Krishnan, Ramayya, Kaplan, David, "On generalized access to WWW-based network of decision support services", Technical Report, The Heinz School, Carnegie Mellon University, 1995; to be presented at the ISDSS conference in Hong Kong, June 1995.

C. ORGANIZATIONAL ASPECTS OF DATA DISSEMINATION

Third-party mechanisms for resolution of disputes in the privacy and information area are examined. These mechanisms include privacy and information clearinghouses, a "Better Data Bureau", a privacy information advocate, a data ombudsman, a privacy mediator, an internal privacy review board, and a data and access protection commission. Applications are made to health care privacy and data access issues.

Faculty working on the project
George Duncan (PI)
Sanda Kaufman, Cleveland State University

Conference Presentations and Papers
George T. Duncan, S. Kaufman, "Who should resolve information and privacy conflicts?: Organizational design for intercessory mechanisms", Technical Report, The Heinz School, Carnegie Mellon University, 1995.

George T. Duncan (1995) Design of intercessory institutions, Organization Sciences Winter Conference, Snowmass, Colorado, January 2-6.

George T. Duncan (1995) Data For Health: Standards For A Health Care Information Infrastructure. A paper commissioned by the AAAS Human Rights Project.

D. A COGNITIVELY GUIDED APPROACH FOR MODEL CONSTRUCTION AND INFORMATION BROKERING

This project is designing and developing a system based on the SOAR architecture to integrate information from multiple data sources. The objective is to implement brokers that can answer queries which require access to multiple data bases.

Faculty working on the project
Ramayya Krishnan (PI)

Graduate student working on the project
Xiaoping Li

Conference presentations and papers
Krishnan, Ramayya, D. Steier, X. Li, "A cognitive approach to model construction and information brokering", technical Report, The Heinz School, Carnegie Mellon University

E. TEAM NEGOTIATION AND DISPUTE RESOLUTION

Bargaining tactics are examined in the context of team negotiation theory and decision analysis. Negotiating teams provide opportunities for role-differentiation strategies, and, in particular, use of such tactics as good cop/bad cop.

Faculty working on the project
George Duncan (PI)
Susan Brodt, Duke University

Conference presentations and papers
Duncan, George T., S. Brodt, "Role-differentiation strategies for negotiating teams: an analysis of 'good cop/bad cop' tactics", technical report, The Heinz School, Carnegie Mellon University This paper was presented as an invited paper to the Annual Meeting of the International Association for Conflict Management last June. Besides the theoretical work on this, they are currently doing an empirical study using videotaped presentations to subjects.

Invited presentation, Strategies for dyadic team negotiations: An analysis of "good cop/bad cop" tactics, Annual Meeting of the International Association for Conflict Management, Eugene, Oregon, June 12-15 (joint paper with Susan E. Brodt).

F. FORECASTING USING BAYESIAN HIERARCHICAL MODELS

"Borrowing Strength from Neighbors" is a promising methodological concept. As implemented in a hierarchical Bayesian model for forecasting, "borrowing strength" leads to statistical tools for effectively using time series data on a related cross-section of units.

Faculty working on the project
George Duncan (PI)
Wilpen Gorr, Carnegie Mellon University

Graduate student working on the project
Janusz Szczypula

This work is the basis for the Ph.D. thesis titled "Adaptive Bayesian pooling methods: comparative study on forecasting small area infant mortality rates"

Conference Presentations and Papers
Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz. Bayesian hierarchical forecasts for dynamic systems: case study on backcasting school district income tax revenues. To appear in New Directions in Spatial Econometrics (edited by Luc Arselin and Raymond Florax)

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz (1993) Bayesian forecasting for seemingly unrelated time series data: application to local government revenue forecasting. Management Science, 39: 275-293.

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz. (1994) Invited paper, Comparative study of cross sectional time series methods for time series with structural changes, The Institute of Management Science/Operations Research Society of America 37th National Meeting, April 25, Boston

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz (1994) Invited presentation, Managing a study of privacy and information issues, University of Maryland/University of Michigan Joint Program in Survey Methodology, College Park, MD, April 27.

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz (1994) Invited presentation, Restricted data versus restricted access: A perspective from "Private Lives and Public Policies," Seminar on New Directions in Statistical Methodology, Council of Professional Associations on Federal Statistics (COPAFS), Bethesda, MD, May 25-26.

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz (1995) Invited presentation, Comparative study of cross-sectional methods for time series with structural changes, International Society for Forecasting, Stockholm, Sweden, June 12-15.

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz, Invited paper, Bayesian shrinkage methods for cross-sectional time series forecasting, 41st North American Meeting, Regional Science Association International, Niagara Falls, Ontario, Canada, November 17-20 (joint paper with Wilpen Gorr and Janusz Szczypula).

Duncan, George T., Gorr, Wilpen, and Szczypula, Janusz (1995) Adaptive Bayesian methods: comparative study on forecasting small area infant mortality rates. Symposium on Statistical Methods, Small Area Statistics in Public Health: Design, Analysis, Graphic and Spatial Methods. Atlanta, GA, January 25-26. Invited for publication in Statistics of Medicine.

2. CONTRIBUTION OF THE PROJECT TO EDUCATION AND HUMAN-RESOURCE DEVELOPMENT

A. WORK WITH PH.D. STUDENTS ON DISSERTATIONS

George Duncan served as major advisor and Ramayya Krishnan as thesis committee member for Sumitra Mukherjee who received his Ph.D. in 1994. Mukherjee was honored with the W. W. Cooper Award for best thesis in management science, Carnegie Mellon University. He currently is Assistant Professor of Computer Information Systems, School of Business and Public Administration, Baruch College, City University of New York.

George Duncan serves as thesis committee co-chair for Janusz Szczypula with expected completion this year. He also is a thesis committee member for Nalini Dayanand, Xianghong Wang, and Dan Zhu, each of whom is expected to complete this year.

Ramayya Krishnan serves as major advisor to Xiaoping Li who is completing a thesis titled "A cognitively guided approach for model construction and information brokering". She is expected to complete her dissertation in the summer, 1995.

B. WORK WITH PH.D. STUDENTS ON PRE-DISSERTATION RESEARCH

George Duncan and Ramayya Krishnan serve as research advisors to second-year doctoral student, Sumit Dutta Chowdhury.

Ramayya Krishnan serves as research advisor to second year doctoral student David Kaplan

C. EDUCATIONAL DEVELOPMENTS

Ramayya Krishnan has developed and is offering a course on Telecommunications Management which covers material on data security and privacy in networks.

George Duncan and Ramayya Krishnan serve on a university-wide task force to examine the potential for distance learning. They also serve on a committee to develop distance learning capabilities for the Heinz School, especially in the area of information systems.

George Duncan is currently developing a course entitled, "Social Issues in Computer Systems".

George Duncan serves as Committee Chair of the American Statistical Association's Committee on Privacy and Confidentiality.