We consider the problem of allocating bandwidth to a set of traffic flows at a statistical multiplexer to provide both quality-of-service (QoS) and resistance to a class of denial- of-service (DoS) attacks. The target application is QoS signaling in a mobile ad hoc network (MANET) environment where the channel bandwidth is variable and the mobile device handles bandwidth requests arriving from multi-hop flows. An admission controller maintains a reserved rate to limit the aggregate traffic rate and to make admission decisions. We analyze the behavior of a rate adjustment scheme based on a Markov Modulated Poisson Process (MMPP) model, which captures the flow-level and burst-level characteristics of variable bit rate traffic. We propose a scheme for adjusting the reserved rate using traffic measurements and an MMPP parameter estimation applied to a reduced MMPP model. Finally, we develop a scheme to estimate parameters for a heuristic rate adjustment scheme that can be executed in real-time. We present numerical results that illustrate the modeling approach and demonstrate the effectiveness of the proposed bandwidth allocation schemes.