Impact of Social Network Structure on Multimedia Fingerprinting Misbehavior Detection and Identification

Abstract

Users in video-sharing social networks actively interact with each other, and it is of critical importance to model user behavior and analyze the impact of human factors on video sharing systems. In video-sharing social networks, users have access to extra resources from their peers, and they also contribute their own resources to help others. Each user wants to maximize his/her own payoff, and they negotiate with each other to achieve fairness and address this conflict. However, some selfish users may cheat to their peers and manipulate the system to maximize their own payoffs, and cheat prevention is a critical requirement in many social networks to stimulate user cooperation. It is of ample importance to design monitoring mechanisms to detect and identify misbehaving users, and to design cheat-proof cooperation stimulation strategies. Using video fingerprinting as an example, this paper analyzes the complex dynamics among colluders during multiuser collusion, and explores possible monitoring mechanisms to detect and identify misbehaving colluders in multiuser collusion. We consider two types of colluder networks: one has a centralized structure with a trusted ringleader, and the other is a distributed peer-structured network. We investigate the impact of network structures on misbehavior detection and identification, propose different selfish colluder identification schemes for different colluder networks, and analyze their performance. We show that the proposed schemes can accurately identify selfish colluders without falsely accusing others even under attacks. We also evaluate their robustness against framing attacks and quantify the maximum number of framing colluders that they can resist.