Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.

Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.

Date of Award


Access Type

Campus Access

Document type


Degree Name

Doctor of Philosophy (PhD)

Degree Program

Computer Science

First Advisor

Brian N. Levine

Second Advisor

Mark D. Corner

Third Advisor

Prashant Shenoy

Subject Categories

Computer Engineering | Computer Sciences | Databases and Information Systems


The concurrent deployment of high-quality wireless networks and large-scale cloud services offers the promise of secure ubiquitous access to seemingly limitless amount of content. However, as users' expectations have grown more demanding, the performance and connectivity failures endemic to the existing networking infrastructure have become more apparent. These problems are in general exacerbated by user mobility.

The work presented in this dissertation demonstrates that performance of services for mobile users is significantly affected by environmental factors that are hard to characterize ahead of deployment. This work includes development and evaluation of large-scale mobile experimentation infrastructures (DOME, GENI) that facilitate longitudinal studies of today's technologically diverse mobile environment over a period of four years. Based on the insights gained from these studies, a mechanism called Spider is presented that provides for efficiently utilizing Wi-Fi deployments in highly mobile scenarios to achieve higher throughput and improved connectivity. This work presents the first in-depth analysis of the performance of attempting concurrent AP connections from highly mobile clients. Spider provides a 400% improvement in throughput and 54% improvement in connectivity over stock Wi-Fi implementations.

The last part of this dissertation demonstrates there are predictable differences among performances of a cellular network in different geographical locations of a town. Consequently, patterns of data transmission between a server on the Internet and a moving cell phone can reveal the geographic travel path of that phone. While the GPS and location-awareness features on phones explicitly share this information, phone users will likely be surprised to learn that disabling these features does not suffice to prevent a remote server from determining their travel path. We showed that a simple HMM-based classifier can discover and exploit features of the geography surrounding possible travel paths to determine the path a phone took, using only data visible at the remote server on the Internet. Having gathered hundreds of traces over a large geographic area, we showed that the HMM-based technique is able to distinguish mobile phones from stationary phones with up to 94.7% accuracy. Routes taken by each mobile phone could be distinguished with up to 75.9% accuracy using the same technique.

This dissertation proposes new tools and techniques for characterization of the impact of the environment on the performance of mobile networks. The concrete set of results and insights gained from this work demonstrates mechanisms for improving connectivity and throughput in highly mobile scenarios while at the same time raising new challenges for maintaining the privacy of mobile users.