On Random Additive Perturbation for Privacy Preserving Data Mining

Title of Thesis: On Random Additive Perturbation for Privacy Preserving Data Mining Author: Souptik Datta, Master of Science, 2004 Thesis directed by: Dr. Hillol Kargupta, Associate Professor Department of Computer Science and Electrical Engineering Privacy is becoming an increasingly important issue in many data mining applications. This has triggered the development of many privacy-preserving data mining techniques. A large fraction of them uses randomized data distortion techniques to mask the data for preserving the privacy of sensitive data. This methodology attempts to hide the sensitive data by randomly modifying the data values often using additive noise. This work presents an analysis of privacy preserving ability of the random value distortion technique in data mining. It points out the possible breach of privacy using random additive perturbation. This work shows that random matrices have ‘predictable’ structures in the spectral domain and it develops a random matrix-based ‘Spectral Filtering Technique’ (SPF) to retrieve original data from the dataset distorted by adding random values. The proposed method works by comparing the spectrum generated from the observed data with that of random matrices. This work presents the theoretical foundation and extensive experimental results to demonstrate that in many cases random data distortion preserves very little data privacy. It presents some direct comparison with previously suggested privacy preserving data mining techniques based on additive random perturbation as well to show the serious breach of privacy. It also explores the possibility of proposed spectral filtering technique on different data types and perturbation methods e.g. discrete data and exclusive or noise. The analytical framework presented in this work points out several possible avenues for the development of new privacy-preserving data mining techniques.