Cross-layer designs for security in wireless ad hoc networks

With the advances in wireless technologies and proliferation of mobile devices, ad hoc networks will play a significant role in enabling future ubiquitous communications. Despite the convenience of forming networks on the fly, without any pre-existing infrastructure, ad hoc networks suffer from serious security concerns. Security in ad hoc networks is challenging due to shared wireless medium, lack of fixed infrastructure, dynamic network topology and resource constrained users. Owing to these differences, conventional security measures used in wired networks are not suitable for wireless networks. To allow interoperability with other IP based networks such as Internet and 3G, existing wireless networks are built on the foundation of OSI and TCP/IP protocol suite. However, the evolving wireless systems seriously question the traditional networking paradigm as these protocol stacks which are architected and implemented in a layered manner are not sufficiently flexible to cope with the dynamics in wireless environments. Cross-layer design has received considerable attention in recent years as the alternative networking paradigm for wireless networks. Through interactions and information exchange among protocol layers, these designs can create wireless architectures with better holistic views of network goals and constraints. The main goal of this dissertation is to design and develop efficient cross-layer architectures and techniques to enhance security and reliability in wireless ad hoc networks.;In the first part of the thesis, we examine various cross-layer architectures to analyze their efficiency in detecting network anomalies. Simulation studies indicate that cross-layer design based on a shared database model has higher system stability and lower overhead. Using this design, we developed CIDS, a cross-layer based intrusion detection system with an objective to provide accurate and reliable misbehavior detection in ad hoc networks. Using multiple linear regression analysis we study the correlation between different variables impacting network performance. This framework helps to identify the cause of a network anomaly and exploits the inter layer interactions to distinguish attacks from genuine network disruptions.;In the second part of the thesis, we leverage the benefits of cross-layer interactions to provide defense against Denial of Service (DoS) attacks. We model jamming attacks at the physical and MAC layers to study DoS behavior, and analyze their impact on network throughput. We develop a cross-layer based measurement driven approach where congestion estimation using physical, MAC and network layer information is used to differentiate jamming and congested scenarios for reliable classification of attacks. Next, we provide a game theoretic framework to formulate jamming as a non cooperative Bayesian game to analyze the interaction between attacker and monitor nodes. We develop hybrid energy efficient detection strategies at the monitor using cross-layer features to achieve the balance in security-energy tradeoff. By estimating the game state, we derive optimal attack and detection strategies.;Finally, we show that the benefits of cross-layer interactions can be extended to security in wireless sensor networks. Tailoring to the needs of the environment, we develop a distributed security architecture (XLSEC) using cross-layer adaptations for layered ZigBee protocol stack targeted towards sensor networks. Here, we incorporate cross-layer interactions within the node as well as among other nodes in the network. In addition, we demonstrate that cross-layer based learning techniques can be used to detect anomalies by correlating the information collected by each sensor node at different protocol layers in a machine learning anomaly detection framework.