User-centrism in wireless networking

This dissertation delves into various aspects of user-centrism in today’s wireless networking landscape. We reconsider the traditional operator/provider-centric view of wireless networking by proposing user-centric solutions to problems along three research dimensions: (i) wireless access, (ii) provision of communication services, and (iii) information collection to be used, among others, for network management purposes. In particular, we first demonstrate cases when users become (micro-)providers of network access themselves. Community wireless networks, where community members share the wireless infrastructure and build wireless mesh networks are one example. We propose a classification of public wireless access schemes and, based on empirical data, discover power-law behavior in their structure. We then focus on the design and implementation of a Wi-Fi sharing protocol on top of resource-constrained, low-cost user devices. Based on this Wi-Fi sharing scheme, we show how user-provided mobile multimedia services can be built in an autonomous, secure, private and decentralized way. We tackle security threats and address legal concerns by proposing a tunneling-based communication scheme with minimal dependence on centralized infrastructures for signaling. We demonstrate the feasibility of secure, user-centric VoIP services to operate on low-cost, off-the-shelf equipment, using a Quality-of-Experience-driven experimental methodology to estimate upper bounds on VoIP capacity in our architecture. Finally, we approach issues of optimizing the operation of the wireless infrastructure in a user-centric way, where discovering the wireless topology (a first step towards reconfiguration and interference mitigation) is crowdsourced to roaming users. We design techniques to counter potential attacks by untrusted users, while we design and implement an architecture based on recently standardized technologies, such as IEEE 802.11k. We derive analytic expressions on the topology discovery accuracy of our scheme, showing that a user-centric scheme can offer more than 2× performance improvement over an infrastructure-centric scheme in realistic scenarios, even in the presence of large ratios of attackers.