Federated Learning and Blockchain-Based Collaborative Framework for Real-Time Wild Life Monitoring

Abstract Effective wildlife monitoring in hilly and rural areas can protect communities and diminish human-wildlife conflicts. A collaborative framework may overcome challenges like inadequate data integrity and security, declining detection accuracy over time, and delays in critical decision-making. The proposed study aims to develop a real-time wildlife monitoring framework using Federated Learning and blockchain to improve conservation strategies. Min-max normalization enhances training data and Elastic Weight Consolidation (EWC) for real-time adaptation. The improvised YOLOv8+EWC enables real-time classification and continual learning and prevents catastrophic forgetting. It also automates actions based on detection results using smart contracts and ensures secure, transparent data management with blockchain. Compared to existing classifiers such as Deep Neural Network, Dense-YOLO4, and WilDect: YOLO, YOLOv8+EWC performs exceptionally well across several metrics, accomplishing an accuracy of 98.91%. Thus, the proposed model enables reliable decision-making by providing accurate, real-time information about wildlife.