Perturbation Methods for Protecting Numerical Data: Evolution and Evaluation

There is a significant need among government agencies and other organizations for tools and techniques that facilitate analysis, dissemination, and sharing of data without compromising privacy and/or confidentiality. Data perturbation techniques are a set of statistical disclosure limitation techniques that try to preserve confidentiality of sensitive numeric data, while retaining the validity of the protected data for analysis. This paper details the evolution of data perturbation techniques, discusses the statistical basis for each technique and compares their effectiveness in terms or protecting the data from statistical disclosure (data security) while retaining analytical validity (data utility).