Hi everyone,

I am attaching the conversation from the assistants list over.

Here is the TL;DR: I want to write my master's thesis on Tor, preferrably on a topic that has to do with Hidden Services and/or Cryptography in Tor.

I have followed George's recommendations and read through some of the sources provided. In the end, several topics seem appealing to me, but before moving on I'd like to get some feedback from you guys on whether you'd consider the topics worth researching or even have some additional ideas.

HSDir tracking: I have taken a look at the idea of PIR ( https://en.wikipedia.org/wiki/Private_information_retrieval) and the problem associated with getting HS descriptiors. I have only looked at the theory of PIR so far and not yet an idea of how this can be accomplished (and to what extend) in practice.

Certificates for HS: I find this topic particularly interesting and have followed the discussion. The general concept seems like a great thing to achieve and it could actually outperform the regular SSL/CA infrastructure stuff as it could remove the need for CAs. Unfortunately, this seems something that is not extensive enough to warrant a whole thesis. If you guys think otherwise, please let me know.

Tor with mix features: Tor has the explicit goal of being a low-latency network. However, there are several protocols where high-latency would be acceptable. I liked the idea of high latency HSes (https://lists.torproject.org/pipermail/tor-dev/2014-November/007818.html). I'd like to know what you think about this idea being viable. It would have the advantage of being very flexible from just a theoretic evaluation down to a real implementation so I could adjust this to my time. But only if this is actually desired so it does not need to stay theoretic. I think it would be very interesting to evaluate whether this can improve or hurt anonymity of low-latency users, as well.

Traffic confirmation attacks: This is here more or less for completeness. I know this topic is open for several years and would be one of the most powerful countermeasures to deploy but unless someone has started on something that I could build upon, I don't see myself coming up with something useful here.

Guard discovery attacks: I have only read roughly what these attacks are. I'd like to know if it would make sense to take a deeper look here, i.e. you think extensive research is needed on that topic.

Improving crypto for HSes: The blog entry on HS (https://blog.torproject.org/blog/hidden-services-need-some-love) vaguely states that crypto for HSes could be improved. However, the article is over a year old and I know the new rend-spec-ng exists, so I'd like to know whether there's anything here to work on. I have a fairly good background on cryptography, so I'd like to help here if help is needed.

Cryptography: There's two proposal ideas, one from 2010 (https://gitweb.torproject.org/torspec.git/blob/HEAD:/proposals/ideas/xxx-crypto-migration.txt) and one from 2011 (https://gitweb.torproject.org/torspec.git/blob/HEAD:/proposals/ideas/xxx-new-crypto-sketch.txt) which builds on that. Has some of this been addressed? Is this still being worked on or just leftover that has already been integrated to the desired level? Would an analysis of the cryptography used in Tor make sense to you, i.e. building on those documents reviewing where and how Tor uses cryptography to secure its operations and evaluating the methods used?

Onion addresses: I took a look at several approaches around censorship-resistant lookups, e.g. the GNS (see George's recommendation below) and Aarown Swartz's proposal on squaring Zooko's triangle by achieving all three properties. I think it would be a cool thing if it were actually possible to improve onion addresses to be human-readable, especially when they get longer by using bigger keys in the future (since 80 bit won't suffice). I don't know if this is actually possible (I see some issues on Aaron's proposal and Dan Kaminsky confirmed them) but working out a scheme that makes handling the names easier for users while not sacrificing the security would help a lot, I think.

This would be the bigger topics I have found on which I could see myself building a thesis. I also stumbled upon smaller research questions (e.g. whether running a bridge/relay is good, bad or doesn't make a difference for anonymity) but none of those warrant a full 6 month thesis so I discarded them for the moment.

If you could take the time to evaluate my ideas and let me know what you think, I'd greatly appreciate that. The hardest thing here as an outsider is to assess the current situation and figure out where work is actually needed and where problems/issues have already been addressed so any help from you guys would really help me.

Thanks in advance & Regards,
Florian Rüchel

P.S.: George:

I'm about to relocate, so my reply will be short! Come and find us in
CCC for more.

Unfortuantely, I don't know what you mean by CCC :(

Ah, I'm also a fan of the FluxFingers team :)

Great! Have played some CTFs for yourselves, then? Are you member of a team?

Thanks for your quick reply, it has helped me a great deal moving forward on this project.

On 12.11.2014 23:15, George Kadianakis wrote: