Data Access, Data Utility, and Disclosure Risk are NOT Always Mutually Exclusive Descriptive Summary to Accompany Presentation at NSF Workshop on Confidentiality May 12-13, 2003 Washington DC

Traditionally, it has been assumed that providing greater data access and/or data utility results in lower privacy/confidentiality. Hence, there is a general feeling that there is no way to provide complete data access and high data utility while also maintaining a low level of disclosure risk. Our research in this area has shown that it is possible to achieve both objectives for confidential numerical data when all relationships between the variables in the data set are monotonic. In this presentation, we illustrate a new masking technique called Data Shuffling that we have developed in order to achieve these objectives for confidential, numerical data. As the name implies, the new technique involves the shuffling the values of the confidential variables without altering relationships among the variables. The new technique provides the agency with the ability to provide complete, unrestricted access to the entire data set, provides the user with data which when analyzed provides the same results as the original data, and assures privacy and confidentiality to individuals and entities from whom the data was gathered.