Research on safety evaluation method of military software

Abstract

It is analyzed that the high reliability demand of ordnance software and the safety evaluation of military software are necessary, and the safety evaluation indicates are confirmed. It is provided that the software safety test and evaluation model utilize the software safety test data getting from the test procedure based on the importance sampling and acceleration testing, to get the evaluation result form the small-sample failure data set using Bayesian estimate estimated method. Take an example of software reliability grown models with nonhomogeneous Poisson process (NHPP) distribution. Firstly the pre-experimental distribution of model parameters are set up utilizing the Bootstrap sampling method, and then the model parameters estimation data are calculated based on Bayesian formula, then the real software accident rate can be calculated integrating the model estimation with the testing acceleration factor. The data case analysis shows that this method is feasible and reasonable.