Hybrid Intrusion Detection for MIL-STD-1553: Integrating Deterministic and Deep Learning Approaches

MIL-STD-1553, widely used in military and aerospace systems, was designed with a focus on safety rather than security, leaving it vulnerable to modern cyber threats. While machine learning and deep learning-based intrusion detection systems (IDS) offer potential solutions, their high computational demands hinder real-time application. This study proposes a hybrid IDS that combines deterministic frame analysis with deep learning techniques to enhance security for MIL-STD-1553 communication. The frame analyzer leverages the protocol's periodic structure to detect anomalies in real-time with minimal computational overhead, while the Long Short-Time Memory (LSTM)-based autoencoder identifies irregularities. Testing on a real avionics platform data, including simulated attacks demonstrates the system's effectiveness. The hybrid IDS outperforms standalone machine learning and LSTM-based approaches, achieving higher accuracy, reduced false positives, offering a practical solution to be used in real-time applications to secure MIL-STD-1553 communication.