DiffT: A Novel Approach for Privacy Preserving Data Analytics

The use of data analytics has led to the creation of automated, accurate, and highly personalized customer experiences across various sectors. However, the use of personal data also exposes individuals to various privacy implications. This research project aims to investigate a technical approach to achieve data analytics without compromising individual privacy. The study in its first phase focused on three primary algorithms, k-anonymization, l diversity, and t-closeness, which are currently used to carry out anonymizations on datasets. It was observed that t-closeness had the least value of re-identification risk. This research aims to propose a novel approach for privacy preserving data analytics named "DiffT". We have enhanced existing techniques and then proposed a novel approach which is combination of both Differential Privacy and t-closeness. The proposed approach DiffT offers input privacy and output privacy control, making it a viable option for privacy-conscious organizations with less demanding computational needs. To benchmark and compare our findings with existing input privacy method Full Homomorphic Encryption, we devise Epsilon (ℇ), as Privacy Loss metric, which is assessed against the total privacy budget. The overall purpose of this research is to reduce the over-reliance on legal and organizational methods, compare the present privacy preserving analytics techniques, and propose a novel intersection of existing models.