Real-Time Secure Cloud Transmission Framework for Agricultural Surveillance Using YOLOv5, AES-256, and HMAC-SHA256

Abstract

The agricultural surveillance systems produce unending floods of delicate visual information that needs to be sent safely to cloud-based infrastructures to be subjected to real-time interpretation and decision-making processes. The current paper introduces a new real-time, lossless, and secure edge-to-cloud transmission model which combines YOLOv5-based event detection, AES-256 symmetric encryption, and HMAC-SHA256 integrity checks and verifies into a single system. The system identifies critical events and encrypts locally and sends only the authenticated data to the cloud thus maintaining confidentiality, integrity and availability. It was evaluated on a custom dataset of 8,000 agricultural images with an average encryption time of 0.17 s/image, decryption time of 0.16 s/image and SSIM of 1.00, which validates the lossless image quality. YOLOv5 model attained 98.5 percent mean average precision (mAP @0.5), which guarantees correct detection prior to encryption. Comparison shows that the suggested approach is faster, more scalable, and more robust in comparison with the existing machine learning and standalone encryption systems. The model provides a scalable architecture of safe, smart farming surveillance, which will form the foundation of future updates pertaining to post-quantum encryption and federated edge learning in precision farming.