AUTONOMOUS ANOMALY DETECTION IN SCADA NETWORKS

This research addresses the growing vulnerability of legacy SCADA networks in critical infrastructure to sophisticated cyber-physical attacks. These systems, often using unsecured protocols like Modbus and DNP3, are ill-protected by traditional, signature-based intrusion detection. This study proposes an autonomous anomaly detection framework leveraging deep learning to identify threats in real-time. By analyzing operational data, models such as the LSTM-Autoencoder learn normal behavioral patterns and flag deviations with high accuracy. The developed system demonstrates superior performance in detecting stealthy attacks like false data injection and command manipulation, significantly reducing detection latency. This data-driven approach provides a proactive security mechanism, enhancing system resilience without costly infrastructure upgrades. It represents a critical shift towards intelligent, adaptive defense for safeguarding essential services against evolving cyber threats, ensuring operational continuity and safety. The framework integrates continuous learning capabilities, enabling it to adapt to new threats and traffic patterns. Experimental evaluation across simulated.