Coordinate Statistics and Gradient Autoencoder Glorot Neural Network Text Based Secured Cryptography

Text-based secure cryptography plays a vital role in web applications, online purchases, banking transactions and so on. Web applications perform operations remotely via Internet to accelerate their processes. However, this improvement in network capabilities comes at the cost of revealing the systems to various attacks. Numerous research works have been developed in the recent years for authentication based data security employing cryptographic methods. In this work a text-based secure cryptography method called, Coordinate Statistics Autoencoder Glorot Neural Network (CS-AGNN) is proposed to maximize information extraction with low execution time and complexity. The CS-AGNN method is split into two sections. First, Coordinate Statistics Filter is applied to the raw secret text message or printed text to remove the noise. Second an autoencoder involving encoder and decoder is designed by utilizing Adam Gradient Autoencoder Glorot Neural Network Text-based Cryptography. This model provides a cryptographic system employing autoencoder-based deep learning model. The model circumvented the requirements of big prime numbers by employing Glorot Initialization synaptic weights of an autoencoder neural network as encryption and decryption keys. The suggested method permits for a high amount of unpredictability employing Adam Gradient function by optimizing XOR results for each round without compromising the network’s overall performance. The results show that the proposed CS-AGNN method outperforms other classical text-based secure cryptography method in terms of execution time (i.e., both encryption and decryption time), key storage cost, throughput and computational complexity. The proposed CS-AGNNmethod has good attributes for text-based cryptography that improves the performance of in terms of key storage cost by 63% and execution time by 59% respectively.