Blockchain and FL-Based Secure Architecture for Enhanced External Intrusion Detection in Smart Farming

Aim:

          The aim of this research is to develop a secure, scalable, and privacy-preserving architecture for smart farming systems by integrating blockchain, federated learning, and ensemble learning techniques to enhance the detection of external intrusions while ensuring data authenticity, confidentiality, and decentralized decision-making.

Abstract:

       Smart farming incorporates advanced digital technologies to enhance agricultural efficiency, productivity, and sustainability. However, the increasing use of interconnected IoT devices also exposes smart farming environments to significant cybersecurity threats, particularly external intrusion attacks that can disrupt operations and compromise sensitive data. Addressing these vulnerabilities requires security solutions that are not only accurate but also scalable, decentralized, and privacy-aware This work proposes a comprehensive and secure architecture that integrates blockchain technology with federated learning to strengthen external intrusion detection in smart farming systems. The architecture allows multiple smart farms to collaboratively train intrusion detection models without exchanging raw data, thereby preserving data privacy and reducing the risk of information leakage. Blockchain plays a crucial role by providing transparent, tamper-proof authentication of all transmitted data, ensuring trust among distributed smart farming units. Sensor readings and authenticated records are securely stored in local base stations, while aggregated results from distributed training are sent to the cloud layer for higher-level analysis, decision-making, and decentralized storage. This layered design enhances the reliability, integrity, and availability of security data across the entire farming network.

Proposed System:

          The proposed system introduces a secure, intelligent, and decentralized smart farming architecture developed using Java Spring Boot for backend operations and Python-based AI prediction for intrusion detection. The main objective of this system is to authenticate sensors, detect suspicious activities, and provide farmers with accurate predictions regarding the security status of their farmland. In this system, farmers begin by completing a registration and login process, after which they are allowed to add their farmland details. During farm registration, multiple sensors such as soil sensors, water sensors, humidity sensors, temperature sensors, and light sensors are mapped to each farmland. These sensor IDs and farm details are securely stored in the cloud. To ensure tamper-proof security, the sensor IDs are also hashed and stored on a blockchain, providing decentralized and immutable verification. When a farmer requests access to any sensor ID, an OTP verification mechanism is triggered. The OTP is sent to the farmer’s registered email. Only upon successful verification is the farmer permitted to view the corresponding sensor ID, ensuring strong multi-factor authentication. The farmland is continuously monitored through connected sensors. Whenever sensor data is transmitted, the system first performs blockchain-based sensor verification to ensure sensor authenticity and prevent spoofing or unauthorized access. Once validated, the sensor data is processed by the Python module (Python Path: Techniques – Federated Learning), where federated learning is used to train intrusion detection models collaboratively across multiple farms without sharing raw data, thereby improving privacy and accuracy. The AI module uses a Random Forest Classifier trained through federated learning, achieving an impressive accuracy of 99.76%. The trained model then performs AI-based prediction to determine whether the farm is under attack or operating normally. The AI prediction results whether the farm is attacked or not attacked—are then updated in the cloud database. Farmers can log in to the system at any time to view the latest prediction results for each of their farmlands. Additionally, they can access complete farm details, sensor information, and the security status provided by the AI model. Overall, the proposed system offers a highly secure, decentralized, and intelligent smart farming solution by combining blockchain-based verification, OTP authentication, cloud storage, and AI-driven intrusion detection. This ensures reliable monitoring, prevents sensor tampering, and provides accurate threat detection for farmers in real time.

Advantages:

High-Level Security Through Multi-Layer Authentication.

           The system combines OTP verification, blockchain-based sensor ID validation, and safe cloud storage, creating a multi-layer security mechanism that prevents unauthorized access, sensor spoofing, or tampering.

Accurate Real-Time Intrusion Detection Using AI

          The integrated Python-based AI model analyzes sensor data continuously, enabling quick and precise detection of unusual activities. Farmers receive instant alerts about attacks, improving response time and reducing potential damage.

Decentralized and Tamper-Proof Sensor Identity Management

          Storing sensor IDs on the blockchain ensures immutability and transparency. This eliminates the risk of manipulation and provides trustable verification for each IoT device connected to the farmland.