Privacy-Preserving Feature Learning on Cloud for Big Data

Abstract

To improve the potency of huge information, feature learning proposed system presents a privacy-protective deep computation model by offloading the pricey operations to the cloud. Privacy issues get to be apparent as a result of the very fact that there’s a major variety of private info by numerous applications within the good town, as an example, sensitive info of Governments, monetary information of assorted organization, or restrictive information of enterprises. As personal information is very important entity, to guard the non-public info, the proposed model uses the BGV a Homomorphic cryptography commit to write in code the non-public info and utilizes cloud servers to execute the high-order back-propagation algorithmic rule on the encrypted information effectively for deep computation model employment. Additionally it also support the secure computation of the activation perform with the BGV cryptography and creates approximate the execution of sigmoid as a polynomial capability. Throughout this originated, solely alone the cryptography operations and then the cryptography operations are executed by the client, whereas all the computation tasks square measure performed on the cloud. Also in contribution, we implement a rule generation method to generate the normal patterns from the predicted outcome by using Apriori algorithm. Experimental results display that system gives greater accurate predicted accuracy than the existing system.