Dynamic Risk Assessment for Civil Aviation Information Systems Using a Multi-Source Deep Learning Framework

Civil Aviation Information Systems (CAIS) are critical for the safety and efficiency of global air transport, but their growing interconnectivity increases exposure to cybersecurity threats, including ransomware, Distributed Denial-of-Service (DDoS) attacks, and human errors. This study introduces the Dynamic Multi-Source Risk Assessment Model (DMSRAM), which integrates Long Short-Term Memory (LSTM) and Fuzzy Neural Networks (FNN) for real-time, adaptive risk quantification in CAIS. Using a 550-day dataset of 13,887 samples from firewall logs, vulnerability reports, botnet logs, and web attack records, DMSRAM dynamically assesses threat probability, vulnerability severity, and asset value. Training results show an RMSE of 0.0167, a detection rate of 92.3%, and an adaptability of 89.4% (indicating stable predictions over time), outperforming the NIST SP 800–30 benchmark (RMSE: 0.0450, detection rate: 75.8%). By capturing temporal dynamics, handling uncertainty, and leveraging multi-source data, DMSRAM offers a robust solution for CAIS cybersecurity. This research provides a comprehensive foundation for data-driven risk management in civil aviation and other critical infrastructures.