Software Reliability Modeling and Prediction of Residual Defects for Embedded Systems

one of the possible approaches to assessing the reliability of real-time embedded systems software is substantiated; the features of the system functioning, criteria and metrics are specified. There are a lot of frameworks based on software reliability growth models (SRGMs) and the Non-Homogeneous Poisson Process (NHPP) for measuring the reliability of software products and predicting residual defects. Numerous SRGM extensions take into account bug correction, debugging workflow, software metrics, programmer qualifications, etc. However, few works are devoted to approaches to evaluation taking into account the features of real-time software. The applicability, goodness of fit, accuracy of the models, and prediction of residual defects are analyzed and compared depending on the initial/input data and the type of real-time tasks to be solved. The novel framework is proposed; the most suitable reliability models included in the framework are briefly considered, in accordance with the different behavior of the risk function over time, and assumptions. The developed framework allows you to choose an appropriate model or set of models to predict the remaining defects in the software. The case study draws on a data set of the results of debugging and functioning of the system for monitoring the state of electrical equipment. The software of such a system implements tasks of mixed criticality with a certain set of periodic and sporadic tasks in real-time.