An Efficient Design and FPGA Implementation of a Hyper-Chaotic and Deep Learning Based PRNG for Image Security Application

Pseudo Random Number Generators (PRNGs) are fundamental to secure communication and cryptographic systems. Traditional chaos-based PRNGs have gained attention due to their sensitivity to small perturbations and inherent unpredictability; however, they are often vulnerable to mathematical attacks due to their deterministic nature. To mitigate this drawback, this paper introduces a novel PRNG design that integrates a deep learning approach to enhance security. A Feed-forward Neural Network (FNN) is trained on four-dimensional (4D) hyperchaotic sequences to capture their complex and chaotic behavior. The training achieved an MSE equal to 1.2585 × 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">−05</sup>• The proposed FNN-based PRNG is implemented on FPGA hardware using VHDL, achieving high efficiency and hardware compatibility. The generated sequences were rigorously evaluated, passing the NIST statistical randomness tests, and were further validated through their application in an image encryption scheme. This implementation meets critical security and real-time performance requirements for cryptographic applications, achieving an operational frequency of 110 MHz.