Privacy preserving classification mining is one of the fast-growing subareas of data mining. The algorithm-related methods of privacy-preserving are designed for particular classification algorithm and couldn�t be used in other classification algorithms. To solve this problem, it proposes a new algorithm-irrelevant privacy protection method based on randomization. This method generates and opens a new data set that is different from the original data set independently as the perturbed data. The perturbed data and the original data have the same distribution. Users get the models of the original data from the perturbed data. Experimental results demonstrate that the classification algorithms can be used on the perturbed data directly. And this method reduces the privacy data disclosure risk more effectively.
|