- tor-dev - lists.torproject.org

Torsocks 2.0.0-rc1 release
by David Goulet 29 Aug '13

29 Aug '13

Greetings everyone! Two months ago I sent an email [1] to the tor-dev mailing list explaining some important issues with torsocks. I proposed myself to do a complete rewrite for the reasons detailed in the email. I've released today the first release candidate for a 2.0.0 version. https://github.com/dgoulet/torsocks Package: https://github.com/dgoulet/torsocks/archive/v2.0.0-rc1.tar.gz I'm asking the community for feedbacks, testing, code review, optimization design and/or anything useful to help improve the project (even correcting the mistakes or improving any comments). The TODO file in the repository contains some needed features but that are not critical to the project for a beta release. Hopefully, with your help, we will be able to replace the current version with a better one, more tested, more hardened and more portable!... and maybe with a more meaningful name :). What is really needed right now is OS X and BSD testers. I did my best to only use POSIX standard and making sure that libc calls are the same on all platforms but I was not able to test it on other OS other than Linux so they might some issues. There is *no* Windows support but again that would be amazing if someone could port it. Again, please contribute and help this very useful tool getting better! I encourage you all to use tor-dev mailing list for any development issues. I'm also on IRC in #tor-dev as "dgoulet". You can also find me on OFTC, freenode and indymedia servers. Big thanks to everyone! Using it is testing it! :) Cheers! David [1] https://lists.torproject.org/pipermail/tor-dev/2013-June/004959.html

4 7

Trac Ticket 1922
by Bao Nguyen 28 Aug '13

28 Aug '13

Providing a fix for #1922, I urge whoever has the authority to decide what folder tor is to parse. I saw /etc/torrc.d/ and /etc/tor.d/ or any others. I am going to make a patch. So which one is it?

1 0

HSDir hash ring modification
by Kang 28 Aug '13

28 Aug '13

Hello. I was reading about hidden services and a thought occurred to me regarding the hash ring used in choosing and determining the HSDirs for a hidden service. As far as I can tell the hash ring is more or less static since a relay's position is determined by their identity key, which doesn't change. I'm also under the impression that the hash ring is only used for calculation of HSDirs of hidden services. I don't have a particular method in mind, but it seems to me that you could use the "time-period" value that is used in calculation of the service's descriptor-id to shuffle the ring. This would cause the ring to be different for each hidden service, and also make its order change periodically. I imagine in particular it would make onion address enumeration attacks more difficult, since an attacker wouldn't just be able to "cast a net" over the ring for all services. Can anybody see any problems or false assumptions with this? Thanks

2 3

Trac Ticket 8291
by Bao Nguyen 28 Aug '13

28 Aug '13

Hi, I am one of the volunteer from Whonix project, a third party application. We want to use tor and modify it to parse the folder /etc/tor.d/ for configuration files. Please take a look at https://trac.torproject.org/projects/tor/ticket/8291 What is the process to get it approved and check in the file. I have modified the file src/or/config.c Who should I talk to? Thanks

3 4

[GSOC] Status report - Tor capabilities
by Cristian-Matei Toader 27 Aug '13

27 Aug '13

Hello tor-dev, Apologies for submitting the report later than planned, but I was on holiday towards the end of last week. I am currently waiting for a code review for the stage 2 of the seccomp filter, which will very likely include a relatively long list of change requests. As a reminder, at the moment there should be one filter which does both syscall and parameter filtering. As a brief update the following changes were made for the past 2 weeks: - various syscall filter changes for linux 32/64 bit - dealt with issues related to the open syscall, due to the fact that we use path filtering and external dependencies happen to use 'open' with strings unprotected by the filter (libevent, getaddrinfo, openssl) - received feedback and added socket filters, which support parameter filtering only for 64 bits, since on 32 they get multiplexed on one syscall (socketcall) - waiting for review on [1] in public branch [2] For stage 3, further privileges will be dropped for different sections of code (eg. the worker threads), and the project will support multiple filters. More details may be found here [1]. Looking forward to some feedback, if you happen to have any! References: [1] https://trac.torproject.org/projects/tor/ticket/9249 [2] https://github.com/cristiantoader/tor -gsoc-capabilities/tree/gsoc-cap-stage2

1 0

Status report - Stream-RTT
by ra 26 Aug '13

26 Aug '13

Hello! Here goes my fourth status report on reducing the RTT of Tor circuits: I continued to evaluate lots of RTT-data of Tor circuits and drew some conclusions from that: *) The RTT measurements of circuits are subject to multiple influences and hence very diverse. *) There is no single distribution that fits them for all circuits. I am currently in the process of doing a demonstration of the improvement the active RTT probing yields. I delayed this report to provide some actual results but the demo is still half-baked.. Best, Robert

1 0

[GSOC 2013] Steganography Browser Extension Status Report
by Hareesan 25 Aug '13

25 Aug '13

Hi, In my previous mail, I was planning to add steganography libraries to the extension. But with the meeting with mentor, we planned to complete the code with proper documentation and to add public key cryptography features. But in my plan, I was able to complete all the UIs and back end framework for all the parts of the extension other than few updates remaining in options page. In cumming weeks, I will be working on completion of code and will add public key cryptography libraries which need some modification from existing libraries. -- Hareesan

1 0

Haskell packages?
by Nikita Karetnikov 23 Aug '13

23 Aug '13

I'd like to improve my Haskell skills. Are there any opportunities? I've been told there is at least one project that uses Haskell, which is not maintained. (For example, this page [1] mentions TorDNSEL, which was replaced by TorBEL.) [1] https://www.torproject.org/getinvolved/volunteer.html.en

5 21

GSOC Status Report - EvilGenius
by Johannes Fürmann 23 Aug '13

23 Aug '13

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello *, During the last two weeks I was primarily busy packing literally everything I own into boxes and unpacking it again, as well as other logistics that have to be done when moving. My equipment as well as my brain and the rest of me should be fully operational again on monday. In spite of the loads of stuff I had to do the last weeks, I still made a few improvement on my projects: * Probe running and logging functionality is now in place * A few improvements on networking: Addresses are now actually set and the corresponding links are pulled up. * Traceroute instrument for debugging purposes. You all have a wonderful weekend ~johannes -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.20 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ iQIcBAEBAgAGBQJSF64/AAoJEAgie5o3Q/JpVN4P/00Zbt3l5H0ASF5KBeUE1/7F jTfyDZzYBKr1dV48V4dExkIdud6dtS6gTeaU9gjlrkH/jr6n81Dg4cxJVEH17n2M wrFSN7QiwVOrD/iPfApJhERIT+h3S503ra2Jejg2nnQTwF86Ifcgvr8gMvKus1b5 ddP8nZQSZqY5e+oZBkVUnjvXW4elc7f/XjRmfs6DmwyvTUGYC8Q+B1+bTMEauQU3 9VlY96HzjTzAKqST/7VihVFwToN0T7KK8pIZeH6O6JN7xcrbX2PfjqvRbPrMgrCE PV82yGVLnzX2jJkIbCFjkZZBTZJKd3X8hZ0A6lftko7SKHJBriSl9lEmxRHtq1pP FQVGAB1x/LEDgcaImupXT9IcnY6BJm2mdZ1Le+CPwY3RJq/CdChdxftGDCPqxl62 6yKOYMIUPxE+I9Cw+LZviFOcRCyH9TxOe918fLJeZ/buPViq+Xf1tw9duBHqq/xr OWTZL2qb5fjcJTG5uomGvwi7fOYq3gTHI/1JFlq9abVpUICrns1nEtevwP7RSb7D LyxIm1zsyIzt+8Qgyxh9uVnOrQY6nZiFmAGy8V3KTxIboAr+Z2booF+EgHwroR/g Sn8bcNng6Ln/cvWsU24OoG53/a4stLo5yNMzFj+p/hU5uvYF9R/4/IDJ6SqqqBzR cEOFb9tzJhCZVD5W9rLb =gX0j -----END PGP SIGNATURE-----

1 0

Global semi-passive adversary: suggestion of using expanders
by Paul-Olivier Dehaye 23 Aug '13

23 Aug '13

Hello, Thank you for working on Tor. I have a suggestion and would appreciate input. Please bear with me as I have a limited understanding of the design of Tor and all the different threats that it is meant to mitigate. Below, a (?) indicates a place where I need some confirmation that my understanding is correct, and N indicates either the number of Tor nodes, the number of end-users, or the amount of traffic (I assume these are all linearly related). As far as I can tell, the main threat by a global passive adversary comes from traffic analysis (?). This attack should become easier as the number of Tor nodes increases (?): Tor uses a clique topology, so the number of edges potentially carrying traffic grows like N^2. A dual way to see this is that not enough mixing can happen around a node for incoming/outgoing edge pairs, bar injecting a huge amount of fake traffic. To compensate, it seems natural to look for a sparse yet highly mixing network topology. Mathematically, those are called expanders [1]. A typical example of a family of expanders would be the Erdos-Renyi model [2], and indeed I have found in the literature suggestions for basing anonymizing protocols on such a model. The analysis in the presence of an active adversary becomes very difficult though. Alternatively, one could use a different method for constructing that expander topology, working "all at once". This comes from recent mathematics research (<= 5 years, certainly not my own, see [3]). The graph is then a Cailey graph [4] in a matrix group (the group is fixed and determined by an approximation to the number of Tor nodes, such as nearest third power of a prime number). In some sense this construction interpolates between mixing chains and Tor, and can be seen as a lot of mixing chains interwoven. In the setting of Tor, constructing the Cailey graph would require making two distributed randomize choices: - a matching of elements of the group to Tor nodes (possibly 2:1 for some Tor nodes) - a small subset of generators for the Cailey graph >From my understanding of security protocols, it should be easy to do these two choices safely and fast, as it amounts to choosing a random element in S_N and filling lots of matrix entries with random elements between 1 and a prime p, with some rejection. Once that is done, the network topology is fully determined, and with very high probability gives an expander. This means that traffic gets mixed up in very few hops. The number of hops needed grows as log N, with a constant that can be mitigated by chosing a large generating set above. This is the only downside I see (apart from difficulty to explain the math behind this): the latency would increase, from 3 in the current protocol to maybe 10 or so. I don't know the details of the behaviour of the constants in the last paragraph, and would appreciate feedback from the list before looking too much into this. Paul Dehaye [1] http://en.wikipedia.org/wiki/Expander_graph [2] http://en.wikipedia.org/wiki/Erd%C5%91s%E2%80%93R%C3%A9nyi_model [3] http://terrytao.wordpress.com/2011/12/02/245b-notes-1-basic-theory-of-expan… 15 and remark below [4] http://en.wikipedia.org/wiki/Cayley_graph

4 4