Below are some brief descriptions of possible Master's and undergraduate thesis topics related to research done by the Distributed Systems Group. Ideally, before starting work on the thesis, you would have completed all or almost all your course requirements and be willing to devote at least one full semester to the project. This is particularly true for Master's theses. We would meet regularly to brainstorm and to discuss progress and problems along the way. For Master's and PhD students (and potentially very strong undergraduate students), if the qualifications and timing are right, we may be able to provide you with funding.

Potential Master's thesis topics

Many of the descriptions below will make more sense if you have taken the graduate courses MDE 518 (Advanced OS) and MDE 519 (Distributed Systems).

• Symbolic execution has been applied/used in computer systems research for a variety of applications including privacy leakage detection in smartphones, bug finding, and others. This thesis shall survey the range of systems applications that have used symbolic execution.
• Survey the existing literature on censorship circumvention systems and propose a new technique or system advancing the state-of-the-art.
• Survey the exisitng literature on privacy of medical data and propose mechanisms for the longterm preservation of large collections of medical data.
• Read and understand the NSDI 2009 paper on ("TrInc: Small Trusted Hardware for Large Distributed Systems) and related works. Read the SOSP 2003 paper on "Preserving Peer Replicas by Rate-Limited Sampled Voting." How would you modify/enhance the LOCKSS preservation protocol if a trusted computing primitive like TrInc were available, particularly for a LOCKSS network with a small number of peers?

• Consider how the LOCKSS protocol might be improved if peers used measures of diversity (e.g., in the implementations of other peers) to influence their opinion poll decisions.

• The behavior of the LOCKSS protocol is highly dependent on the topology of the friends lists and evolving reference lists held by participating peers. This topic focuses on studying the behavior of the protocol under varying topologies of these lists. Some prior theoretical analysis has already been done on the system and would be provided as a starting point. This topic is appropriate for a theoretically-inclined student as well as a more systems-inclined student, depending on how the student approaches it. Work would include implementing simulations to verify the accuracy of the current model, testing the model under varying topology configurations, and measuring the resilience to Sybil and other attacks given particular topologies. Ideally, we'd find two students, one with analytical interests and one with systems interests to work in parallel here.

• Implement a distributed web site watchdog tool. Given a set of sites, monitor the sites for changes, track and preserve the changes, then display the changes in a user-friendly manner. Lots of challenges here including: synchronizing web crawls so that system components agree on the ordering of web page versions, figuring out how often to crawl a web site, and presenting changes to the user in an easy-to-read manner. There is enough meat in this topic for at least two Master's and/or PhD theses.

• Read the 2009 ACM TON paper on PAN called "A Directory Service for Perspective Access Networks" by Goodell et al. (see DS group publications web page). Read the CoNEXT 2009 paper on content-centric networking called "Networking Named Content" and related papers (See: www.ccnx.org).

  Find the connections/synergies between PAN and content-centric networking. Could PAN be a basis on which to perform content-centric networking? Note: this topic and the next are for students with networking expertise who are particularly interested in next-generation Internet architectures.

• The PAN architecture was in many ways influenced by the desire to incorporate it in a widely-deployed system (Tor) which provides: 1) a way to support unmodified applications, 2) a source-routing framework, 3) a tunneling mechanism and 4) anonymity. Could the design of PAN be simplified or made more lightweight if we did not have the constraint to run it on Tor?

• Related to the above topic, build a simple web-page perspective sharing system. This would be for the specific purpose of web page sharing -- not the general goal of PAN, which is to enable sharing of any kind of online resource and requires lots more bells and whistles.

• Survey privacy-enhancing technologies (PETS) for web browsing and propose your own or improve upon the state of the art. Specifically, we might be interested in monitoring traffic that comes in and out of our browsers as we surf the web so we can track what input (e.g., info on a form or cookies) we have provided to which sites and when.

• Design and build a privacy-preserving social network. First, you would need to survey privacy issues in social networks, including any related works such as Diaspora. Start with a google search for "privacy preserving social network" and see what you get.

• Often students have interesting ideas of their own. Feel free to propose your own thesis topic and we can discuss.

Potential Undergraduate Thesis Topics

The following are possible undergraduate thesis topics. These are difficult as they require the student to work, for the most part, independently and to take initiative in answering questions on his/her own. Each involves building a system or application from scratch.

• Evaluate different approaches to network communication in Java (TCP/IP,Selectors,Threads,SSL).
• Implement an asynchronous, event-driven efficient communication infrastructure in C++, allowing the application to be ignorant of specific network infrastructure, while using high level services like "send", "sendall", "setTimeout".
• Convert a distributed protocol's (Bracha consensus) Java implementation, to C++.
• Compare two DOM trees and produce a third one with the differences highlighted.
• Produce an add-on to a contemporary browser that logs all outgoing communication; then present it in a meaningful way.
• Implement a peer-to-peer protocol using WebRTC.
• Implement an Android client for an e-voting application.
• Acronym generator: More often than I care for, I find myself looking for a sensible acronym to represent the name of a research project. Needless to say, this is a boring and time-consuming task. There are some tools on the Internet that suggest possible acronyms given the name of a project, but these are not satisfactory as they typically just spit out rote acronyms in a brute-force manner (e.g., pick first letter of each word), etc and they don't perform any ranking of the returned results or attempt to return results that are also real words. The purpose of this project is to build a web portal that accepts phrases (e.g., a name like "Distributed ApplIcations and SYstems"), computes possible acronyms that can be derived from the phrase, finds which combinations are real words and returns a ranking of the acronyms, with real words (e.g. DAISY) receiving higher ranking than acronyms that are just a bunch of letters that don't mean anything. Number of students: two.
• Build a Facebook application that does something interesting. This project involves figuring out on your own how to build a Facebook application and designing an interesting application. Part of the difficulty here is in writing a design document that describes your potential application and convinces the prof this is an interesting enough application. Implement your application and make it available. See if you can figure out how many facebook users use your application and even better, if you can trace the line of friends from you to a Facebook user that used your application.

• Build a Facebook app that helps users "fight censorship of web content". (See prof for more info).

• Build a "binary bomb" assignment that could serve as a future syspro project. That is, see if you can replicate this Harvard class assignment.

• Build a system where you connect to any node you want of a set of nodes and send a request for a url to that node. That node then opens via wget a connection to retrieve the web page at the url of interest and sends back the web page it retrieves. Extension: implement and deploy this application on Planetlab so that you may choose any node and see what a particular web page looks like from its perspective.

• Often students have interesting ideas of their own. Feel free to propose your own thesis topic and we can discuss.

All of the above undergraduate projects require motivation and drive on the student's part. If you have the appetite to work independently and take initiative to seek answers to questions on your own and would like to work on one of these topics, please email your cv and a transcript to mema@di.uoa.gr.