Covering designs for efficient de-identification of health records

In Canada, research ethics boards vary widely in the way they interpret risk to patients when their personal information is being re-used for research purposes. However, many research ethics boards will permit information to be shared without consent when they believe there is a low risk of re-identification. PARAT is Privacy Analytics’ de-identification software tool. It automatically performs generalization and suppression on cross-sectional, longitudinal, and geospatial data. The program measures the re-identification risk in the database, and can simulate attacks under a variety of different assumptions. The development of smaller covering designs and their incorporation into the PARAT de-identification tool will allow Privacy Analytics to handle larger datasets than are currently possible. As such, Privacy Analytics will be able to market PARAT to bigger health organizations with more valuable data sets then would have otherwise been feasible. Our research plan is to adapt successful approaches used in other covering problems, such as Aickelin’s algorithm for the set covering problem, to the problem of computing covering designs.

Faculty Supervisor:

Dr. Liam Peyton

Student:

Andrew Baker

Partner:

Privacy Analytics

Discipline:

Engineering

Sector:

Information and communications technologies

University:

University of Ottawa