Data Privacy and Prediction Using Neural Network and Homomorphic Encryption
Nilesh D. Navghare, D.B. Kulkarni · 2018
To enhance the correctness of learning result multiple parties can associate through conducting Back-Propagation neural network learning jointly on the combination of their individual data sets. In this process, none of the parties reveal his/her private data to the world. Traditional schemes support a combined learning but it is limited to two parties only. So there to perform a combined learning if a data is partitioned arbitrarily for more than two parties. This paper resolves this open hassle via using a strength of cloud computing. In this scheme, every single party creates encoded text and transfers the encoded text to the cloud. Cloud then performs ample no of operations on encoded text with the help of learning schemes without knowing the original private data. While securely transferring operations towards the cloud, the cost incurred during transfer of operations should be minimal and independent to parties involved. To perform diverse operations on encoded text, “homomorphic asymmetric encryption” scheme is used for multiple parties.