Multi-Encryptions System Based on Autoencoder Deep Learning Network

Abstract

Encryption is a very important stage in data security. It allows only authorized people to see private data after decryption. This paper proposes Multi-Encryptions System (MES) for encrypting and decrypting images based on the Autoencoder Deep Learning (ADL) network. Four layers of ADL was used to encrypt spcific images multiple times and reconstruct the entered data again. These are the input layer, encryption layer 1, encryption layer 2 and decryption layer. The proposed MES provides high security as it has the ability to generate and utilize multiple encrypted information from original data. To measurement the reliability of this work different statistical calculations are applied. The results are promissing where significant encryption-decryption performances can be obtained.