Artificial intelligence has used in many scientific fields fundamentally, but its use in data protection, particularly in encryption and decryption, has been a supportive and auxiliary tool. This research presents a mechanism for using artificial intelligence to perform fully hybrid encryption, eliminating the RSA and AES algorithms entirely and replacing them with multiple neural networks that perform both encryption and decryption.
The mechanism for combining the RSA and AES algorithms was studied, taking into account the cipher lengths. The 12-bit AES algorithm is used fifteen times, compared to the 248-bit RSA algorithm once, to obtain the cipher. This is achieved after segmenting the m-message into mi sub-messages. Forty distinct neural network architectures underwent training for English plaintext encryption and decryption operations. Each network had three characters as input and three ciphers as output. This number of neural networks is used to solve the problem of the number of elements in the training database, as 40 networks were used for encryption alone, and another 40 networks were used for decryption, along with the segmentation and pooling operations.
The proposed model was programmed in the MATLAB GUI environment, and the encryption and decryption times required to perform the proposed model were determined and compared with the execution times of the RSA and AES algorithms. Evaluation results showed that the proposed model outperforms RSA in time of encryption while approaching AES encryption speeds. The results also demonstrated that this model outperforms both the RSA and AES algorithms in decryption, demonstrating its fast performance.