AI-Driven Cyber Threat Detection in Hybrid SCADA and Space-based IoT Systems: A Lightweight Anomaly Detection Approach

As critical infrastructure and remote industrial operations increasingly rely on interconnected control systems and satellite-enabled IoT devices, cyber threats targeting Supervisory Control and Data Acquisition (SCADA) and space-based IoT environments pose significant risks. These hybrid systems suffer from legacy protocols, limited computational resources, high-latency satellite channels, and heterogeneous communication paths—all of which hinder traditional intrusion detection systems (IDS). This paper proposes a unified AIdriven cybersecurity framework designed to detect anomalies and cyberattacks in hybrid SCADA–space-IoT environments. Leveraging machine learning and deep learning architectures such as Autoencoders, LSTM, and CNN-LSTM hybrid models, the system analyses telemetry sequences, command packets, and communication patterns in real time. A combination of real-world datasets (NASA SCaN, ESA OPS-SAT) and synthetic attack logs (generated via NS-3 and MiniCPS) is used to evaluate detection performance. Experimental results demonstrate a detection accuracy of up to 97.5% with low latency, making the system suitable for resource-constrained field devices and satellite-driven industrial networks. The proposed work sets a foundation for next-generation cyber-physical security across both terrestrial and space-borne infrastructures.