PRIVACY PRESERVATION FOR DISTANCE BASED DATA MINING IN DISTRIBUTED DATA

Data collected from vast, diverse devices for data mining purposes must be preserved from privacy and security issues. This research aims to improve the privacy preservation for distributed data in data mining without each participant knowing the other participant's data and address the concern of information loss at every data processing stage. We designed an algorithm for preserving privacy for distributed data mining by combining existing privacy-preserving techniques with the orthogonal random transformation matrix. We introduced the homomorphic encryption process to avoid a third party's involvement in the encryption computation process. We re-formulated the algorithm and recalculated the distance of the inner product for any two records; therefore, we could apply the privacy-preserving model to the traditional data mining processes. Our results achieved the same accuracy for both the traditional and privacy-preserving models, showing that our approach was practical and efficient. However, our results of the time of our privacy-preserving model exhibited a higher computational time but which was, however, a trade-off for a more secure system.