Evaluation of time and resource efficiency of application of dynamic bayesian networks models for organization of testing of web applications by phasing

Abstract

The web application test procedure is a complex process that includes separate test modules and information models that establish the relationship between test modules and their elements, taking into account the specifics of detecting certain classes of software errors. The current dynamics of the development and operation of software tools and systems for creation of effective automated testing methods, pre-assigned for high-quality, in-depth analysis and search for software errors. In addition to the further development and modification of known approaches and protocols, it was necessary to create new methods, models and algorithms that would allow statistical analysis of the results of testing, forecasting and timely localization of the most critical areas of the programs. The study examines the use of dynamic Bayesian networks to simulate the process of testing web applications with MT phasing. The paper proposes dynamic Bayesian models for testing the main groups of errors in web applications, allocated in accordance with the classifications OWASP and MITRE. Probability graphical models based on dynamic Bayesian networks allow you to present testing in the form of a stochastic process with a fixed set of time states, where each subsequent state is evaluated taking into account the processing of the results of previous ones. Evaluation of time and resource efficiency of application of dynamic Bayesian networks for organization of procedure of web-applications testing by fuzzing is carried out on the special metrics and indicators. Special metrics are proposed for constructing the structure of dynamic Bayesian networks, tuning parameters and implementing probabilistic output in the tasks of prediction, filtering and smoothing.