Securing IoT Data Transmission with Deep Learning Approach to Steganography with Convolutional Neural Networks and Generative Adversarial Networks

In the evolving Internet of Things (IoT) landscape, network data security is a critical concern due to vulnerabilities like weak passwords, insecure networks, and outdated components. This project tackles these challenges by introducing an innovative encryption method that conceals IoT network data within image files. By addressing key vulnerabilities, this approach enhances data transmission security. The project employs Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs), integrated into traditional steganography techniques to increase payload capacity and se-cure data transfer. Testing results demonstrate a 0.4000 increase in accuracy (achieving 90.04%), a significant reduction in loss by 0.4525, and improved image quality with a Peak Signal-to-Noise Ratio (PSNR) increase of 23.8752, showcasing the model's robustness in real-time encoding and decoding. The ultimate objective is to provide a comprehensive solution to strengthen IoT data security in the face of emerging threats.