一个基于“恶意对抗”的公平交互机制的研究

由于多代理交互环境中检测机制、仲裁机制的有限性，合法代理将注定要遭受损失。针对上述现象，该文提出了一个恶意对抗机制（Malice-Antagonized Interaction Mechanism,MAIM)，即在该机制下，交互代理双方都允许表现出恶意行为，以期实现交互的公平性，为了分析MAIM机制的可行性，文中应用游戏理论对MAIM在几种典型的游戏中实现交互公平性的情况进行了分析和证明。最终得出结论：在完善信息的纯策略游戏中，MAIM无法实现交互的公平性，而在完美信息的混合策略游戏、不完美信息的顺序游戏和同时决策游戏中，能实现交互的公平性。从而从理论上证明了恶意对抗机制的可行性。

A Malice-Antagonized Interaction Mechanism for Fair Interaction

With outstanding advantages in constructing distributed system, the research and the application on the mobile agents become more and more attractive. However, at the same time, security problem always is a key obstacle for further extending the application of mobile agent, especially the security problem on mobile agent self-protecting. This paper tries to make an exploration in this field with the focus on the malicious behaviors in the mobile agent interaction. Because of the malice of interact counterpart and the limitation of detection mechanism, the normal or legal agent will have the loss in the interaction. Based on the activities in the physical society, normal agent being malicious will make the interaction become fair. Therefore, this paper presents our Malice-Antagonized Interaction Mechanism(MAIM)that allows agents to become. In order to analyze the viability of MAIM in the computable space, we apply game theory and by that we make a series of analysis on that mechanism in several typical types of game, i.e. pure strategy game with perfect information, mixed strategy game with perfect information, sequential strategy game with non-perfect information, and simultaneous strategy game. And after that, we get the conclusions that in the pure strategy game, it is impossible to get the fair interaction under our MAIM model, however in other MAIM models, it is viable.