Learning from Privacy Preserved Encrypted Data on Cloud Through Supervised and Unsupervised Machine Learning
Ahmad Neyaz Khan, Ming Yu Fan, Asad Malik, Raheel Ahmed Memon · 2019 2nd International Conference on Computing, Mathematics and Engineering Technologies (iCoMET) · 2019
With the advent of new technologies and the ever increasing use of Cloud in nearly every sphere of our day to day life, the data owner using cloud is still less confident. This lack of trust is obvious as the data owner entrusts the data with a third party which stores, manages and processes the data. Whatever be the level of security, there is always some loop hole for the misusers. Various machine learning techniques have been in use to learn from the data available for analysis and to use the results accordingly for benefits. Homomorphic secure multi-party computation (SMC) or homomorphic encryption (HE) encryption schemes have been one means to securely process the data on cloud while preserving the privacy of the data. In this work we have tried to investigate both supervised and unsupervised machine learning capability through neural networks over encrypted data from a semantically secure cryptosystem based on Homomorphic properties. This work will provide a base for the machine learning performance over the data on cloud whose privacy is claimed to be preserved using Homomorphic encryption. The findings are supported with experimental results.