Abstract: To solve the problem in attack-defense strategy research of complex network, an analysis method based on rough Bayesian game was proposed. By introducing rough set theory into object Petri net, the authors defined Attack-Defense Confrontation Model (A-DCM), which divided domain strategy set into equivalence class and accordingly put forward the characteristic strategy set. Rough Attack-Defense Game Model (RA-DGM) and utility function were presented, and the Bayesian equilibrium solution algorithm was given, so that the maximal attack-defense strategy set could be acquired. The analysis method was suitable for attack and defense research of complex network, which could reduce strategy space scale effectively. The experimental results verify the correctness and performance of the model, and demonstrate that the attack-defense strategy analysis method based on the model is more reasonable and effective.

Key words: network attack-defense, Bayesian game, rough Petri net, Attack-Defense Confrontation Model (A-DCM), maximal attack-defense strategy

摘要： 为了解决复杂网络上攻防策略研究中存在的问题，提出了基于粗糙贝叶斯攻防博弈模型的分析方法。扩展传统对象Petri网并在其节点域和变迁域上引入粗糙集理论，定义了网络攻防对峙模型。通过在论域攻防策略集上划分等价类，提出了特征攻击策略集的提取方法。在此基础上定义了粗糙攻防博弈模型(RA-DGM)以及攻防效用函数，并给出了该模型的贝叶斯均衡以及最大攻防策略集的求解算法。该分析方法能够有效缩减博弈分析中策略空间的规模，适合于复杂网络攻防行为研究。实例说明了该模型的正确性及其对冗余信息的处理性能，基于该模型的攻防策略分析方法也更为合理、有效。

关键词: 网络攻防, 贝叶斯博弈, 粗糙Petri网, 攻防对峙模型, 最大攻防策略