Novel Solution Approach for Multi-Objective Attack-Defense Cyber Games With Unknown Utilities of the Opponent

A nontraditional game-theoretic approach has been recently suggested for solving Multi-Objective Games (MOGs). This paper aims to highlight the applicability and advantage of this approach to cyber security. For this purpose, we describe a new attack-defence game-theoretic model. It involves a zero-sum MOG in which the players are undecided about their objective preferences. As demonstrated here, solving such a unique type of game results in a set of rationalizable strategies wherein each strategy is associated with a set of performance vectors. Set-based strategy selection techniques are introduced and demonstrated. In addition, existing evolutionary algorithms to finding the set of rationalizable strategies are discussed. The suggested cyber-game model and the nontraditional solution approach are demonstrated using a simple academic example. Based on the demonstrated case, we conclude that the described approach opens up new possibilities for analyzing cyber defense strategies. Namely, it allows tradeoff considerations by eliminating the need to predefine a utility function for each player. Moreover, based on a time-complexity analysis, we also conclude that future algorithms for solving real-life multiobjective cyber games, by the proposed method, should be adjusted for parallel processing. Finally, in view of practical needs for cyber security, this paper takes a broader view and discusses how the development of game-based computational tools for cyber defence, such as proposed here, may benefit not just from evolutionary computation methods but also from other computational intelligence methods, such as fuzzy inference systems and artificial neural networks.