Energy-Efficient Chaos Encryption Scheme and Image Processing in Wireless Multimedia Sensor Networks Using Swin Transformer-Aided YOLOv8 Model

Wireless media sensor networks (WMSNs) are widely used for real-time data collection, monitoring and flexible deployment, yet challenges remain regarding security, efficiency and energy consumption due to open communication, large data volumes and resource limitations. This paper presents an energy-efficient encryption and image processing scheme for WMSNs using deep learning. Raw images from the DOTA-v2 dataset undergo preprocessing steps such as resizing and denoising, where the K_TF technique enhances image quality. The SwinT-YOLOv8 model is employed for dynamic object detection, followed by a chaos-based lightweight image encryption (CLIE) scheme that ensures secure image encryption. This multi-stage encryption uses random values from Henon’s chaotic map, Brownian motion and Chen’s chaotic system to reduce pixel correlation and enhance security. The integrated approach facilitates secure communication and image exchange among network nodes. Implemented in MATLAB, the proposed system achieves 99.25% accuracy, a detection time of 60 s, a network lifetime of 60% and energy efficiency of 80.37%, along with improved performance across various key parameters.