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.

Author ORCID Identifier


Open Access Dissertation

Document Type


Degree Name

Doctor of Philosophy (PhD)

Degree Program

Computer Science

Year Degree Awarded


Month Degree Awarded


First Advisor

Amir Houmansadr


The Internet plays a crucial role in today's social and political movements by facilitating the free circulation of speech, information, and ideas; democracy and human rights throughout the world critically depend on preserving and bolstering the Internet's openness. Consequently, repressive regimes, totalitarian governments, and corrupt corporations regulate, monitor, and restrict the access to the Internet, which is broadly known as Internet \emph{censorship}. Most countries are improving the internet infrastructures, as a result they can implement more advanced censoring techniques. Also with the advancements in the application of machine learning techniques for network traffic analysis have enabled the more sophisticated Internet censorship. In this thesis, We take a close look at the main pillars of internet censorship, we will introduce new defense and attacks in the internet censorship literature.

Internet censorship techniques investigate users’ communications and they can decide to interrupt a connection to prevent a user from communicating with a specific entity. Traffic analysis is one of the main techniques used to infer information from internet communications. One of the major challenges to traffic analysis mechanisms is scaling the techniques to today's exploding volumes of network traffic, i.e., they impose high storage, communications, and computation overheads. We aim at addressing this scalability issue by introducing a new direction for traffic analysis, which we call \emph{compressive traffic analysis}. Moreover, we show that, unfortunately, traffic analysis attacks can be conducted on Anonymity systems with drastically higher accuracies than before by leveraging emerging learning mechanisms. We particularly design a system, called \deepcorr, that outperforms the state-of-the-art by significant margins in correlating network connections. \deepcorr leverages an advanced deep learning architecture to \emph{learn} a flow correlation function tailored to complex networks. Also to be able to analyze the weakness of such approaches we show that an adversary can defeat deep neural network based traffic analysis techniques by applying statistically undetectable \emph{adversarial perturbations} on the patterns of live network traffic.

We also design techniques to circumvent internet censorship. Decoy routing is an emerging approach for censorship circumvention in which circumvention is implemented with help from a number of volunteer Internet autonomous systems, called decoy ASes. We propose a new architecture for decoy routing that, by design, is significantly stronger to rerouting attacks compared to \emph{all} previous designs. Unlike previous designs, our new architecture operates decoy routers only on the downstream traffic of the censored users; therefore we call it \emph{downstream-only} decoy routing. As we demonstrate through Internet-scale BGP simulations, downstream-only decoy routing offers significantly stronger resistance to rerouting attacks, which is intuitively because a (censoring) ISP has much less control on the downstream BGP routes of its traffic. Then, we propose to use game theoretic approaches to model the arms races between the censors and the censorship circumvention tools. This will allow us to analyze the effect of different parameters or censoring behaviors on the performance of censorship circumvention tools. We apply our methods on two fundamental problems in internet censorship.

Finally, to bring our ideas to practice, we designed a new censorship circumvention tool called \name. \name aims at increasing the collateral damage of censorship by employing a ``mass'' of normal Internet users, from both censored and uncensored areas, to serve as circumvention proxies.