Dissertation for obtaining a scientific degree PHD in the speciality 123 – "Computer Engineering". – National technical university "Kharkiv polytechnic institute". – Kharkiv, 2023.
The subject of research – software security enhancement method.
The object of research – software security process.
The dissertation work is devoted to the solution of the current scientific and technical problem of developing a method of improving the security of software, taking into account the possibilities of synthesis of technologies of automated software security testing and deep machine learning. The study of the software life cycle and testing processes accompanying this cycle was performed using graph theory (GERT modeling).
The development and research of the method of automated safety testing was carried out using the method of deep learning with reinforcement. Improvement of the method of evaluating the effectiveness of the developed method was carried out using the method of dynamics of averages.
The assessment of the reliability of theoretical and practical results was carried out using the provisions of probability theory and mathematical statistics.
The scientific novelty of the obtained results is due to the theoretical generalization and a new solution of an important scientific and technical problem, consisting in the development of a method for improving the security of software based on penetration testing technologies. The following scientific results have been obtained. – for the first time, a method of automated intrusion testing using the Shodan search engine, the MulVal network security analysis platform, and CVE software vulnerability data has been developed to input and build realistic attack and validation scenarios for deep learning with reinforcement technology. This allowed to generate an attack tree for various training procedures, to optimize the corresponding scenarios of automatic software security testing, and thus increase the efficiency of the software security process; – improved mathematical model of the process of testing for penetration into computer systems, different from the known capabilities of testing the security of specialized information platforms of computer systems, which allowed to estimate the probability of testing time for penetration in a given interval; – the mathematical model of the process of testing for penetration into computer systems was further developed. A distinctive feature of this model is the use of the Erlang distribution as the main one in the mathematical formalization of the processes of transition from state to state. This allowed, on the one hand, to unify the mathematical model and present the testing process at a higher level of the testing hierarchy, on the other hand, to simplify it.
The practical significance of the obtained results is to adapt the software testing process to the increased security requirements and capabilities of testing automation tools, using deep learning technologies with reinforcement. The practical significance of the obtained results is as follows. – a set of mathematical models of the penetration testing process in computer systems using the GERT network modeling approach simplified the penetration testing scheme by 1.7 times, taking into account possible changes in procedures (including the addition of new procedures and services) to estimate probabilistictemporal characteristics and possibilities of its scaling at increase in volume and complexity of the solved problems; – the synthesis of the main components of the method of automatic penetration testing has increased the efficiency of the software security process (reduce the relative damage at all stages of the software life cycle to 6 times).
The results of the dissertation are implemented and used in the activities of the company "Line Up", the Research Center for Forensic Examination on Intellectual Property, and are also used in the educational process of the National Technical University "Kharkiv Polytechnic Institute".
The relevance of the thesis topic is justified in the introduction, the main goal and task of the work are formulated, and the scientific novelty and practical value of the obtained results are presented.The first section is devoted to the analysis and comparative studies of software penetration testing methods. In the second section, the penetration testing process mathematical models complex. In the third section, automated penetration testing method using deep machine learning technology are developed. The fourth section is devoted to the study of the efficiency of the software security improving method and substantiation of practical recommendations for its use.
Key words: software, automated testing, security testing, deep machine learning, fuzzy GERT model, cyber threat, software vulnerability, software security mismatch.
