Applying Machine Learning for Run-time Bug Detection in Aviation Software
Hu Huang, Samuel Z. Guyer, Jason H. Rife · 2016
Modern commercial aircraft are heavily dependent on large software systems for many of their essential functions. However, bugs may cause software failure and endanger the lives of passengers and crew. Software development processes in aviation emphasize near exhaustive testing but not all bugs can be found. Additionally, current bug detection methods, including assertion-based methods, are not suitable for bugs that appear during run-time as they produce complex relationships between program variables. In this paper, we investigated a machine learning approach for detecting bugs at run-time. Our preliminary results show that our approach can detect the presence of injected software bugs with high specificity. This technique shows promise in meeting the high standards of aviation reliability.