ABSTRACT
This paper presents a Machine Learning-guided sampling strategy to predict high-vulnerability fault injection targets in RTOS, achieving 100% fault coverage with 10x fewer injections.
PAPER · PDF
Loading PDF…
Key findings
ML-guided sampling reduces campaign size by 10x while maintaining equivalent fault coverage.
The approach achieves 100% fault coverage with only 666 injections compared to 6,660 for exhaustive campaigns.
Execution time is reduced by 4x, enabling integration into CI/CD pipelines.
Limitations & open questions
The study focuses on four RTOS platforms, limiting the generalizability of the findings.
Further research is needed to expand the approach to other types of faults and RTOS configurations.