Methodical approach for evaluation of the number undetected natural semantics defects with the required degree of confidence in the verification of space vehicles on-board computer systems software

The approach for evaluation the number of undetected natural semantics defects in the program is developed and presented in the framework of creating software verification method for space vehicles on-board computer systems. Method is created to reduce the total cost of verification that satisfied the tasks of the verification process. Detection of this defects type is performed within the static analysis of the program source code. It is based on monitoring compliance with the principle of dimensional uniformity of expressions. The estimation of the defects number is performed according by the Mills reliability model, which belongs to the statistical models class with deliberate injection of defects into the software. The defects injection procedure is demonstrated on the representation of the program via sets, taking into account the specific natural semantics defects characteristic of the source code, which affect the adequacy of the procedure performed. The required evaluation degree of confidence of undetected defects is achieved by calculating the number of detectable injected defects expected value. This statistic characteristic received from the selection of the required size, which depends on the statistical characteristics obtained from the initial selection and the given values of the confidence probability and the confidence interval. The obtained approach can be used in research of the programs source code within the process of software verification for space vehicles on-board computer systems, since it is based on well-known mathematical relationships and meets the requirements of normative documents in this field.