Detecting MAC layer misbehavior and rate adaptation in IEEE 802.11 networks: Modeling and SPRT algorithms

Abstract

As the commercial success of the IEEE 802:11 protocol has made wireless infrastructure widely deployed, user organizations are increasingly eager to discourage new threats to their networks as well as enhance the network performance. An emerging threat to these networks is the intentional misbehavior at the medium access control (MAC) layer. As wireless stations are usually under the control of individual users, a selfish or greedy user may try to gain more bandwidth at the expense of other cooperating users. As an extreme, when a station constantly occupies the wireless medium, a denial-of-service (DoS) attack is launched and the misbehaving station becomes an attacker. Efficient mechanisms are needed to detect such stations misbehaving at the MAC layer. On the other hand, advances in wireless communications research have led to key enabling technologies, and next generation wireless networks are expected to provide reliable, high-bandwidth, and on demand services. As the flexibility of adapting physical layer transmission modes is crucial for realizing broadband wireless communications, multiple transmission rates at the physical (PHY) layer are provided in the standard of the 802:11 protocol. Although this multi-rate capability is critical to the system performance, no dynamic rate adaptation (or control) mechanism is specified in the standards.

In this research, we address two key issues critical to the performance of the 802:11 networks. We first address threats posed by DoS attacks against the MAC layer protocol. In our approach, an analytical model depicting greedy behaviors of attackers is developed. In particular, we present a Markov chain model to describe the stochastic trend of a station's behavior cheating on the MAC layer protocol and derive the distribution of inter-delivery times between two consecutive successful transmissions. Based on this inter-delivery time distribution model, we develop a set of algorithms to detect misbehaving stations using the sequential probability ratio test (SPRT) and analyze their performances through simulations. We then focus on the PHY layer's multi-rate capability and develop a rate adaptation scheme that incorporates the SPRT and uses a feedback mechanism to amend rate adaptation decisions. Experimental results obtained through simulations show that our algorithms significantly improve the performance over the existing solutions.

Contributions of this dissertation lie in two areas. Firstly, we develop analytical models of various characteristics in the 802:11 network that can be extended to a wider application domain. Secondly, we propose solutions to the two key problems from a statistical perspective fully exploiting the SPRT technique that significantly improve the performance of the 802:11 network over the existing ones.