Cybersecurity Considerations for AI/ML Assisted Student Flight Training

This paper describes the cyber security concerns and methods for mitigating risk in a near real-time AI/ML assisted student flight training project undertaken at Embry-Riddle Aeronautical University in Prescott, Arizona. The goals of the project, architecture for data acquisition and the processing of data for use in training, assessment and maintenance are described. The cybersecurity risks and attack surface are highlighted with the most concerning areas emphasized. Methods for mitigating the risks from different attack vectors are provided with the express understanding that new approaches for breach of confidentiality, integrity, accessibility and availability may not have been considered in detail. In conclusion, the next steps and plans for follow on efforts are discussed offering problems for both undergraduate and graduate research projects. The motivation for this effort is clear, with the demand for highly skilled pilots growing worldwide, methods that can decrease the training time and at the same time increase the quality of graduating pilots by incorporating timely feedback in a non-subjective manner via AI/ML assistance is worth investigating.