Game Theoretic Modeling of Malicious Users in Collaborative Networks

摘要

If a network is to operate successfully, its users need to collaborate. Collaboration takes the form of following a network protocol and involves some resource expenditure on the part of the user. Therefore, users cannot automatically be expected to follow the protocol if they are not forced to. The situation is exacerbated by the presence of malicious users whose objective is to damage the network and increase the cost incurred by the legitimate users. The legitimate users are, at least initially, unaware of the type (legitimate or malicious) of the other users. Our contribution is a model for the strategic interaction of legitimate and malicious users as described above. The model is based on repeated graphical games with incomplete information. We describe and analyze two specific instantiations, aiming to demonstrate the modeliquests expressive power and tractability. The main benefit we see from using game theory for this essentially security problem is the ability to bound the damage caused by the malicious users.