Asymmetric Encryption Method Proposed for Arabic Letters Using Artificial Neural Networks

Abstract

Asymmetric encryption algorithms suffer the problem of high execution time. This paper presents a proposed methodology to perform encryption and decryption of messages written in Arabic, using artificial intelligence represented by artificial neural networks based on the RSA algorithm. The method is based on dividing the message into partial messages. The arrays of weights and biases in the neural network layers are considered the encryption or decryption key, according to their function. All steps of the proposed method and how it works were shown and used in the MATLAB environment to design a system for encrypting and decrypting messages written in Arabic. The results proved the effectiveness of the proposed methodology and its superiority over the RSA algorithm in terms of execution time for both encryption and decryption. The encryption time in the proposed algorithm is close to the time of decryption, unlike what it was in the RSA algorithm, where the encryption time was relatively much greater than the decryption time. The proposed method was tested on four text files of 50KB, 100KB, 150KB, and 200KB, over two hundred iterations. The average improvement in execution time was 1,330ms and 4,497.5ms for encoding and decoding, respectively. KEYWORDS Data protection, decryption, execution time, security, RSA algorithm, performance