Protecting Privacy in case based reasoning by disordered PCA on one class data
Wei Lü, Tian-Rong Zhong, Xun Liao, Rui Wu · 2010
Protecting Privacy has attracted more and more attention in data mining. Case based reasoning(CBR) is very important task in data mining. This paper presents method that protects the privacy by disordered principal component analysis(PCA) on one class data. In order to be ensure the security of the CBR, we first disorder the PCA to select the principal component confusedly. Further we transform the sensitive attribution into principal component space using disordered PCA, thus the sensitive attributes are encrypted and protected. Because the PCA method is disordered, this algorithm is very secure. In addition, PCA can keep the main character of dataset, so the precision change of CBR after encryption can be controlled in a small scope. The experiment show that if we select appropriate parameters, then nearest neighbors of every point may be high consistent. The present algorithm can guarantee that the security and the precision both achieve the requirements.