A Hybrid Machine Learning Based Intrusion Detection System for MIL-STD-1553

Abstract

MIL-STD-1553 is a communication standard developed by the US Department of Defense in 1975, primarily utilized in military aircraft, ground vehicles, and spacecraft. Due to its dual-redundant data bus structure, high reliability, and low error rate, it finds applications in safety-critical systems. Despite being considered secure at the time of its development, MIL-STD-1553-based systems have gradually become vulnerable over time, presenting easy targets for attackers. In this research, we propose the utilization of a new hybrid method based on machine learning and natural language processing for MIL-STD-1553, aiming to perform anomaly-based intrusion detection. In doing so, we employ Stochastic Gradient Descent and BERT algorithms, previously unused for intrusion detection in the MIL-STD-1553 system. The proposed system was experimentally evaluated against cyber-attacks. We observed that the hybrid intrusion detection system provides satisfactory results in detecting intrusions on MIL-STD-1553 data bus. Overall, experimental results show that the proposed system may be used to detect intrusions on MIL-STD-1553 based communications.

Keywords

• mil-std-1553
• intrusion detection
• avionics
• machine learning
• bert