Traceability of Executable Codes Using Neural Networks

摘要

Traceability of codes refers to the mapping between equivalent codes written in different languages - including high-level and low-level programming languages. In the field of Legal Metrology, it is critical to guarantee that the software embedded in a meter corresponds to a version that was previously approved by the Legal Metrology Authority. In this paper, we propose a novel approach for correlating source and object codes using artificial neural networks. Our approach correlates the source code with the object code by feeding the neural network with logical flow characteristics of such codes. Any incidence of false positives is obviously a critical issue for software evaluation purposes. Our evaluation using real code examples shows a correspondence around 90% for the traceability of the executable codes with very low rate of false positives.