Blockchain-enabled federated learning-based privacy preservation framework for secure IoT in precision agriculture

The aim of this paper is to establish a secure and privacy preserved IoT communication in precision agriculture. For achieving security and privacy, federated learning system have been deployed on blockchain ecosystem to classify IoT communication attacks in precision agriculture. This paper has utilized recent 'CICIoT2023' database to automate identification of prominent cyber-attacks in IoT. Sharing data between devices raised privacy concerns but without sharing data knowledge also getting limited for classification of diverse attacks. So, we have deployed federated learning ecosystem over Ethereum block chain to achieve collaborative learning with privacy preserving communication. In methodology, initially recent dataset about cyber-attacks classification have been collected, pre-processed and distributed for multiple devices. The integration of the Ethereum blockchain with IPFS decentralized file storage for transmitting the learning model from client device to server and vice versa enhances the overall security and trust of the system. Initially basic machine learning algorithms have been employed in standard single machine environment to establish benchmark results. Then a deep neural network has been deployed in blockchain based federated learning environment to analyse the outcome using identical and non-identical data distributions. In results significant outcomes have been achieved in terms of privacy and security with high accuracy, precision, recall, etc., while training deep neural network. This paper has worked for number of subset data classifications to propose and analyze overall view for securing IoT communication from cyber-attacks in precision agriculture.