A Framework for High-Accuracy Privacy-Preserving Mining

Shipra Agrawal, Jayant R. Haritsa

Abstract: To preserve client privacy in the data mining process, a variety of techniques based on random perturbation of data records have been proposed recently. In this paper, we present a generalized matrix-theoretic model of random perturbation, which facilitates a systematic approach to the design of perturbation mechanisms for privacy-preserving mining. Specifically, we demonstrate that (a) the prior techniques differ only in their settings for the model parameters, and (b) through appropriate choice of parameter settings, we can derive new perturbation techniques that provide highly accurate mining results even under strict privacy guarantees. We also propose a novel perturbation mechanism wherein the model parameters are themselves characterized as random variables, and demonstrate that this feature provides significant improvements in privacy at a very marginal cost in accuracy. While our model is valid for random-perturbation-based privacy-preserving mining in general, we specifically evaluate its utility here with regard to frequent-itemset mining on a variety of real datasets. The experimental results indicate that our mechanisms incur substantially lower identity and support errors as compared to the prior techniques.