Resource allocation and pricing mechanisms for wireless networks with malicious users

In this thesis, we propose a framework for designing decentralized pricing and centralized auction mechanisms in the presence and absence of malicious users. First, we design pricing mechanism for efficient power allocation in the uplink of a single cell Multi Carrier Code Division Multiple Access (MC-CDMA) system with strategic and price taking users. The mechanism is extended to sum power constraint over the users and carriers. Additionally, prices are designed for different designer objectives such as revenue maximization and energy minimization. We also consider multihop heterogeneous wireless networks and design appropriate pricing mechanisms for efficient joint power and rate allocation. An iterative algorithm for pricing mechanism is proposed and its convergence is proven. We use the pricing mechanism framework to design a location privacy mechanism for mobile commerce with entropy utility functions and a budget constraint. A regression learning method is used by the designer to learn the utility functions of users from their actions. This is in contrast to direct mechanisms where the designer asks the users to report their utility functions. Next, a centralized auction mechanism for networks with interference coupled utilities and price anticipating users is proposed and its efficiency is proven. A new modeling of malicious user utility function is proposed and a metric Price of Malice (PoM) for quantifying the effect of malicious users is defined. The PoM and related metrics for pricing and auction mechanisms are obtained. Then both the pricing and auction mechanisms are extended for countering the malicious users using differentiated pricing. Next, we relax the assumption that the users and the designer know the nature of users and design Bayesian mechanisms. By comparing to the complete information case, the conditions under which the uncertainty about the nature of the users is beneficial for the regular users and the designer, are obtained. The probability of a user being malicious is constructed using learning methods and the differentiated pricing is implemented using this probabilistic information. Then pricing mechanisms with malicious users are extended to a case where each user submits a Quality of Service(QoS) requirement.