NonHomogenous Anonymization Approach Using Association Rule Mining for Preserving Privacy

Statistics reveals that with the advancement of technology, data sharing has become crucial for the purpose of research and analysis. Many organizations share their data with third party venders to gain information regarding the relevant and meaningful patterns that are hidden among the vast amount of raw data. But as the data needs to be shared with the third party, maintaining privacy of the individual becomes a challenging task. Thus, it becomes critical to share data in such a way so that privacy of the individual records is not hampered. This requirement has escalated the demand of research in a relatively new field that is privacy preserving data mining. Many techniques has been studied and developed related to this field which are based on either central server or distributed server. Some of these techniques include K-Anonymity, I-diversity, cryptographic method, randomization etc. Based on k-anonymity framework, nonhomogenous approach of generalization has already been proposed earlier. This approach of nonhomogenous anonymization provides privacy to the identity of individual records but suffers high privacy risk factor. The objective of the present work is to extend the existing approach of nonhomogenous generalization using association rule mining. Experimental results show that proposed technique performs better than existing technique based on two evaluation parameters - i.e. data disclosure Risk factor and execution time.