Smart Energy-Efficient Encryption for Wireless Multimedia Sensor Networks Using Deep Learning

Wireless multimedia sensor networks (WMSNs) have gained considerable attention across various applications due to their capabilities for real-time multimedia data collection, efficient monitoring, and flexible deployment. Despite advancements, challenges persist in ensuring security, optimizing efficiency, and minimizing energy consumption due to the open remote medium, large volumes of multimedia, and inherent resource constraints in WMSNs. This paper introduces an innovative energy-efficient protection model for WMSNs, leveraging advanced deep learning techniques. The model utilizes a lightweight Tiny YOLO-v7 framework to dynamically identify objects within captured images, thereby reducing the need to transmit superfluous data. Moreover, the model combines the lightweight Speck cipher for the encryption of detected objects with a scrambling method that permutes and shuffles all image pixels. An effective key management scheme is also integrated to secure communication and image exchange among nodes within the network. By restricting encryption and transmission to sensitive images containing foreign objects, the proposed model significantly reduces operational overhead. The experimental results showcase the effectiveness of the proposed model in reducing node power consumption by approximately 49% compared to conventional methods, which encrypt and transmit all generated images. Furthermore, the model demonstrates a significant 50% improvement in extending network lifetime compared to related encryption solutions. The security analysis substantiates the model's resistance against diverse attacks, ensuring compliance with the stringent security requirements of WMSNs. Furthermore, the model exhibits strong potential for real-time applications in dynamic monitoring and secure environments.