Supporting secure and transparent mobility in wireless local-area networks
Wireless Local Area Networks (WLANs) are experiencing unprecedented growth as the last mile connectivity solution. Mobility is an important feature of any wireless communication system. Handoffs are a crucial link level functionality that enable a mobile user to stay connected to a wireless network by switching the data connection from one base station or access point to another. Conceptually the handoff process can be subdivided into two phases: (i) Discovery---wherein the client searches for APs in vicinity and (ii) Authentication---the client authenticates to an AP selected from the discovery phase.
The handoff procedure recommended by the IEEE 802.11 standard and closely implemented by various wireless vendors is an intrusive and a brute-force approach. My testbed based study of these algorithms showed that they incur high latencies varying between 400ms to 1.3 seconds depending on the security settings in effect. Such inefficient handoff mechanisms can have a detrimental impact on applications, especially synchronous multimedia connections such as Voice over IP.
In my dissertation, I have proposed and evaluated the notion of locality among APs induced by user mobility patterns. A relation is created among APs which captures this locality in a graph theoretic manner called neighbor graphs---a distributed structure that autonomously captures such locality. Based on this, I have designed and evaluated efficient mechanisms to address the two different phases of this handoff process. Through a rigorous testbed based implementation, I have demonstrated the viability of the concept of mobility induced locality through good performance improvements. Through extensive simulations I have studied the performance of proposed handoff mechanisms over various different topologies. This work has shown that a topological structure which captures the locality relationship among APs is fundamental to designing mechanisms that make user mobility transparent from the higher layers of the networking stack.