Preventing and encrypting large data and perform deep learning computations on big data

Abstract

To improve the efficiency of large data feature learning, the paper proposes a privacy protective deep
computation model by offloading the expensive operations to the cloud. Privacy concerns become evident as results of
their area unit AN outsize vary of private data by varied applications inside the wise city, like sensitive data of
governments or proprietary information of enterprises. To safeguard the personal data, the planned model uses the BGV
secret writing theme to inscribe the personal data and employs cloud servers to perform the high-order back-propagation
method on the encrypted data with efficiency for deep computation model employment. What's additional, the planned
theme approximates the Sigmoid operate as a polynomial operate to support the secure computation of the activation
operate with the BGV secret writing. In our theme, only the key writing operations and conjointly the cryptography
operations area unit performed by the patron whereas all the computation tasks area unit performed on the cloud.
Experimental results show that our theme is improved by almost about 2.5 times inside the employment efficiency
compared to the standard deep computation model whereas not revealing the personal data using the cloud computing
still as ten nodes. Lots of considerably, our theme is extremely scalable by employing a ton of cloud servers that's
particularly acceptable for large data