EFFECTIVENESS EVALUATION OF STATIC ANALYSIS FOR SEARCHING NATURAL SEMANTIC DEFECTS OF PROGRAM OBJECTS
Dmitry Viktorov, Roman Zhidkov
Abstract
The paper evaluates the effectiveness of static analysis for searching a new type of functional defects called program objects natural semantic defects in terms of completeness and accuracy on the basis of the probability theory mathematical apparatus. Calculation of performance indicators is carried out by graph modeling of the defects localization variants in the software code constructions for various groups of operations (arithmetic, assignments, and comparisons). The possible state of the analyzer is modeled for each terminal situation on the basis of a priori known checking construction natural semantic result value statement (true trigger, conditionally true trigger, skip, and norm). The obtained dependencies allow to determine the feasibility of using static analysis to find considered defect’s type on the basis of statistical characteristics of software code