Homomorphic Encryption Schemes Using Nested Matrices

With the increasing necessity to secure data stored in different cloud platforms and transmitted through the internet, many encryption algorithms are developed to protect data from attacks. Homomorphic encryption (HE) which is used to allow performing arithmetic appertains on encrypted data without the need to decrypt is widely used to encrypt data in the cloud platform and privacy preservation for machine learning training. A recently proposed HE scheme uses nested matrices to encrypt and decrypt any field numbers of data so it can be used with machine learning and artificial intelligence applications. In this work, we propose an enhancement to the model which will increase the security and the randomness of the model and make it more reliable to be used with machine learning models. We also present a comparative analysis showing the performance of the originally suggested scheme and the improved scheme proposed in this work.