A chaotic based approach for privacy preserving data mining applications with multilevel trust

Privacy preserving data mining addresses the issues related to maintaining the data privacy while publishing the data. The privacy of the exact data is maintained by publishing the perturbed copies of the data. Traditionally the data is perturbed by mixing the Gaussian noise data. The level perturbation is depends upon the trust (the more trusted a data miner can access the less perturbed copy of the data) on the party for which the data is required to generate. In practical situations the same data is perturbed for different level of trusts (for different uses) or same level of trusts, such situations arise the threat on security of data that may cause estimation of accurate data (especially when multiple copies of data mixed with Gaussian noise are available) from these copies by advance computational algorithm like Linear Least Squares Error (LLSE). This paper presents a nonlinear system based chaotic signal generator for perturbation of original data the complex characteristics of the chaotic signal makes it difficult for the estimators to estimate original data because the estimators works on the known noise probability distribution function (PDF). The simulation result shows that the proposed algorithm gives higher estimation error when compared with traditional algorithms.