In this work, we study a Markov model of cyber threats that act on a computer system. Within the framework of the model the computer system is considered as a system with failures and recoveries by analogy with models of reliability theory. To estimate functionally-temporal properties of the system we introduce a parameter called the lifetime of the system and defined as the number of transitions of the corresponding Markov chain until the first hit to the final state. Since this random variable plays an important role at evaluating a security level of the computer system, we investigate in detail its random distribution for the case of mutually exclusive cyber threats; in particular, we derive explicit analytical formulae for numerical characteristics of its distribution: expected value and dispersion. Then we generalize substantially the Markov model dropping the assumption that cyber threats acting on the system are mutually exclusive. This modification leads to an extended Markov chain that has (at least qualitatively) the same structure as the original chain. This fact allowed to generalize the above analytical results for the expected value and dispersion of the lifetime to the case of non-mutually exclusive cyber threats. At the end of the work the Markov model for non-mutually exclusive cyber threats is used to state a problem of finding an optimal configuration of security remedies in a given cyber threat space. It is essential that the formulated optimization problems belong to the class of non-linear discrete (Boolean) programming problems. Finally, we consider an example that illustrate the solution of the problem on selecting the optimal set of security remedies for a computer system.

The problem of selecting an optimal collection of security remedies against a specified set of cyber threats is relevant as far as there is a wide range of different cyber security solutions in modern IT industry. The aim of this work is to formulate an optimization problem for selecting information security remedies using a Markov cyber threat model and to analyze the possibility of solving the problem by the method of sequential analysis of variants. The set of solutions for standard approaches is limited by admissible indicators of the economic damage, and the corresponding restriction by means of a functional and temporal characteristic of the model called its average lifetime is defined. The explicit analytical formula for the average lifetime of an information system expressed in terms of the original parameters of the model is obtained, these parameters being the probabilities of threat occurrence and probabilities of their eliminations by security remedies. The possibility of solving our optimization problem by the method of sequential analysis of variants is analyzed. The program in C++ is developed to experimentally compare the effectiveness of this method compared with the "brute-force" method.

No abstract