Published Paper Details:

Data privacy is major concern in cloud computing. Most of the methods preserve the sensitive data like address, phone number, etc. and fewer methods to preserve quasi identifiers. In anonymization and randomization there is information loss during reverse process of preservation. I have worked to check the information loss problem by using K-mean clustering. Database of employee is considered. Three quasi identifiers are preserved during experiment. In gender attribute preservation, age attribute is taken because gender is Boolean. The whole age record is divided into two clusters and the first and the second cluster centroid are assigned to male and female respectively. In case of age preservation, 3 clusters are formed. Initially the difference between maximum and minimum age is calculated and if that difference is small then consider it as a noise, otherwise calculate the remainder of maximum and the minimum centroid and if its value is small, then take it as a noise and merge it in the age data. If both conditions are not applicable then take the remainder of maximum cluster with 10 (constant), and that becomes the additive noise. To preserve the area code, find the difference between maximum and minimum centroid and then multiply it to the original zip code data. So, the privacy of attributes is preserved. During reverse process the age noise is subtracted from preserved age data and the altered area code is divided with noise and original data is received. So the information loss is null in the proposed method.