In this paper we propose an extremum seeking approach based on evolutionary game theory, to solve the dynamic bandwidth allocation problem in wireless networks. The algorithm proposed aims to maximize a general utility function associated to the network, defined as the sum of the individual utilities of the agents (users) that belong to the network. Due to the complex time-varying nature of the available bandwidth, the quality of the links across the population of agents, and the parameters and structure of the utilities of the agents, a non-model based control is required. In this context, the extremum seeking algorithm proposed allows the optimal on-line bandwidth allocation, based only on the individual measurements of the utility function of each agent. We show how the extremum seeking control converges to a neighborhood of the optimal allocation, maximizing the general utility function of the network. This function can be externally manipulated by a network manager aiming to achieve a tradeoff between a “throughput” and a “fairness” behavior in the resource allocation solution.