一种信息不确定系统的模糊安全博弈模型

Ambiguous Game Based on Minimax Regret in Uncertain Information System

面对信息技术广泛运用的今天,信息安全问题开始受到人们的广泛关注.现有的最新工作大多使用贝叶斯博弈的方法来解决.贝叶斯博弈可以应对参与者类型信息不完全的情况,但现实生活中,信息不仅不完全而且是不确定的,即不知道对手类型精确的分布概率.为解决这个问题,扩展了贝叶斯博弈,提出极小极大模糊博弈模型(ABGMR).该模型中使用证据理论来描述信息的模糊性;引入可接受收益的极小极大遗憾原则,在收益可接受的情况下最小化最大遗憾值,避免了出现无法接受的收益和过度悲观的情况.极小极大模糊博弈模型适用于多种攻击者多个防御者的情况,并且攻击者和防御者不需先观察对手的策略,可以同时采取行动.最后将ABGMR运用在信息安全领域资源分配场景中,并与已有博弈算法DOBSS进行实验对比,证明了其有效性和最终决策的最大遗憾值的优越性.

As information technology is widely used,the issue of information security has become a rising concern.Most of the latest available researches use Bayesian game to solve problems.Bayesian game model can handle the case where the information about players’ type is incomplete.However,in real life,the information is not only incomplete but also ambiguous,i.e.,a player may not know the precise probability distribution over potential opponent’s type.To deal this issue,this paper introduces a new game model named the Ambiguous Game（ABGMR）,which extends the Bayesian game model by incorporating Dempster-Shafer Theory that describes the ambiguity and a principle of acceptable costs of minimax regret that avoids the overly pessimistic approach and the potential for unaffordable losses.This model is applicable for the situation with multifarious attackers’ types and various defenders.Attackers and defenders actually take their actions simultaneously.Finally,the model is illustrated in Information Security Domain and contrasts with existing game algorithms,which proves the validity and superiority of ABGMR in maximum regret of final strategies.