Enhancing cybersecurity in Agriculture 4.0: A high-performance hybrid deep learning-based framework for DDoS attack detection

Abstract

Industry 4.0 technologies are transforming agriculture, moving towards Agriculture 4.0: i.e., a new era focused on enhancing productivity and sustainability through advancements such as Internet of Things (IoT), Artificial Intelligence (AI), fog and cloud computing. Devices equipped with IoT technology continuously gather real-time data on soil quality, crop health, and equipment functionality, which is then analyzed via fog and cloud computing to streamline farming operations and improve agricultural efficiency. Although these advancements enhance productivity, they also pose considerable cybersecurity threats, especially in terms of Distributed Denial of Service (DDoS) attacks, which can jeopardize the availability and reliability of essential systems and critical infrastructures. This paper presents a deep learning-driven security framework aimed at mitigating these vulnerabilities in Agriculture 4.0. We propose a hybrid Intrusion Detection System (IDS) integrating a deep-Autoencoder (dAE) for binary classification and a Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) for multiclass clustering. Our framework, exploiting real-world data from the CIC-DDoS2019 dataset to detect DDOS attacks, evaluates autoencoder models alongside HDBSCAN, with each technique tested in three configurations. This combined approach demonstrates effective threat detection and classification capabilities, achieving accuracy levels exceeding 98%, thus enhancing the cybersecurity of agriculture 4.0, promoting robust, data-informed, and efficient farming practices while aligning with Sustainable Development Goals (SDGs) concerning industrial innovation and resilience.