tor-dev

tor-dev@lists.torproject.org

3581 discussions

Re: [tor-dev] Introducing CollecTor
by Karsten Loesing 08 Jun '14

08 Jun '14

[Replying on list with Damian's permission.] On 06/06/14 17:21, Damian Johnson wrote: >> I'm currently extending metrics-lib to fetch descriptors from CollecTor. >> Maybe Damian extends Stem to do the same (or accepts patches for that >> purpose). > > Probably obvious, but what's the advantage of this over fetching new > descriptors from authorities and mirrors? What kind of CollecTor > methods did you have in mind? If you're only interested in new descriptors, then the authorities and mirrors are the place to go. But if you care about having the full history of descriptors, then you should go to CollecTor. Imagine that your tool might break for a few hours. There's no way to get the missing data from the directory authorities or mirrors anymore. For applications like ExoneraTor, where you need to store every single consensus, this can make the difference. The same holds for Onionoo and the metrics website, though it's less critical to miss a few hours. Also, if you only care about relay descriptors, then you should be fine asking directory authorities and mirrors. But sanitized bridge descriptors, exit lists, bridge pool assignments, and Torperf measurements are only available from CollecTor. Note that there's also a disadvantage of fetching from CollecTor: it's a single point of failure. Now, we could add mirrors, but it would be even better to set up a second CollecTor instance (ideally run by another person) in parallel to the current first instance, and have the two exchange data. Future work, on my list. I guess the above should be added to the CollecTor homepage. Do you have any suggestions where you would have expected to read about this, or do you have further points to add? I'll add something next week. All the best, Karsten

1 0

[GSoC] Consensus Diffs - First report
by Daniel Martí 08 Jun '14

08 Jun '14

Hello everyone, This is my first status report regarding my GSoC project, Implementing consensus diffs. We weekly meetings with Nick and Sebastian, my mentors, every Wednesday at 14h UTC, if you want to take a closer look at the project's progress or want to participate. In the application timeline I specified that I would spend the first weeks coding the server side part, which is mostly just the generation of diffs. And that's what I've been doing thus far. I'm a bit behind on schedule due to exams, but I'm not too far off. I first implemented a dynamic algorithm, quadratic in both time and space, which seemed to work fine. Of course it didn't work for consensuses, since that would mean an allocation of nearly 3GB. So the next thing I did was switching to Hirschberg's algorithm, which still takes quadratic time but only takes linear space due to a divide-and-conquer approach. But that still leaves us at a horrible 16 seconds of runtime to generate a regular microdescriptor consensus diff, whereas 'diff -d' only takes a fifth of a second. That is largely due to the fact that we're not taking advantage of the consensus properties yet. So what I will do next is improve the algorithm, which should make it very close to linear in time. The idea is to navigate the two consensuses, looking at the digests that will already be sorted for us. That's part of the 'r' line for descriptor consensuses, and the 'm' line for microdescriptor consensuses. This should take linear time. Now, the part of Hirschberg's algorithm that makes it quadratic in time is finding 'k', the column at which the problem can be split in half wihtout losing precision. This is well explained here: http://wordaligned.org/articles/longest-common-subsequence But once we are navigating the nodes ordered by digests, we can avoid having to calculate 'k'. Given that we are at any pair of matching digests, we can assume that this is already a point 'k' because all the nodes that are before the digest on the first consensus can only be found before the digest on the second consensus, and the same goes for later digests. So the final algorithm should travel both consensuses in linear time, and use our quadratic time algorithm on the small chunks between these matching digests. The quadratic time algorithm would at worst be used for maybe a couple hundred lines, a size at which it still performs fairly well. You may be wondering why I went with an O(MN) algorithm when I could have gone with the more common O(ND) algorithm, which I understand is the one used in the GNU diffutils. The main reason is the use-case. O(ND) performs worse than O(MN) when the differences are large, because then D=2N. In our case, when we are running the algorithm it's because nodes disappeared, appeared or simply to find changes in nodes that are still there. So it will be very common that we find diffs D that are very close in size to N or M. Plus, as far as I understand, the divide-and-conquer algorithm is pretty clean and easy to understand while the O(ND) algorithm by Myers would probably be harder to implement in a clean way. I'm in no way an expert in diff/lcs algorithms, so suggestions and comments are very welcome. -- Daniel Martí - mvdan(a)mvdan.cc - http://mvdan.cc/ PGP: A9DA 13CD F7A1 4ACD D3DE E530 F4CA FFDB 4348 041C

1 1

DetecTor Review
by Damian Johnson 07 Jun '14

07 Jun '14

Hi Kai. I just took a quick peek at your DetecTor project (http://detector.io/). It looks like the project has been inactive for a while so just some quick thoughts in case it sparks back up. > log_setting = "notice file " + data_dir + os.sep + "tor.log" The following might be nicer... log_setting = "notice file %s%stor.log" % (data_dir, os.sep) Same goes for most instances of string concatenation, actually. > 'StrictExitNodes': str(1), Interesting, never seen this before. Should be the following instead. 'StrictExitNodes': '1', > def stop_tor(control_port): > tor_pid = stem.util.system.get_pid_by_port(control_port) > if tor_pid: > os.kill(tor_pid, signal.SIGTERM) Certainly a fine approach. Would you like to avoid having Tor outlive this python process? If so then that's as easy as including 'take_ownership = True' in your launch_tor_with_config() call. If not then you can use the process object returned by launch_tor_with_config() to kill it as well... tor_process = stem.process.launch_tor_with_config() ... do stuff... tor_process.kill() Also this method has tabs mixed with spaces. I would highly recommend strictly using spaces - python is whitespace dependent, and occasional tabs can make things a lot more confusing. > def is_tor_running(control_port): > return bool(stem.util.system.get_pid_by_port(control_port)) I believe you can use the poll() method of the process object to make this a little less platform dependent. > if os.path.isdir(dir_name) != True: You never need to compare with booleans in an if statement. This is equivalent to... if not os.path.isdir(dir_name): > config.readfp(open('sphere_control.config')) Minor thing but this doesn't properly close the file handle. It's probably done on its own during GC, but it would be more proper to do... with open('sphere_control.config') as config_file: config.readfp(config_file) Honestly I'm really not sure what the following code is doing. I suspect there's a lot of low hanging fruit for improving this (for instance using "', '.join(countries_list[i])" instead of tracking a 'countries_string') but first step would be some comments explaining what this is trying to achieve. Cheers! -Damian

1 0

GSoC : Rewrite Tor Weather (Status Update 1)
by Sreenatha Bhatlapenumarthi 07 Jun '14

07 Jun '14

Hi all, So here is my first update on the work regarding weather rewrite. 1. (#12085) Develop python wrapper for Onionoo services. Made an initial python wrapper to be used for fetching Onionoo documents. It currently supports request-response caching to decrease the latency(picked up from OnionPy). 2. (#11956) Wrote a simple script for searching for relays matching user's query. This functionality is to be implemented in the rewritten weather as well. Modified it to use the new Onionoo wrapper instead of 'urllib' module for fetching response from Onionoo. 3. (#9889) Rewrote the previous tshirt script, making use of the new Onionoo wrapper. Future work 1. Add and expand the features provided by the Onionoo wrapper. Currently in the process of adding utility methods for backend scripts that use Onionoo data. 2. Test the Onionoo wrapper 3. Write the remaining backend scripts(hourly, daily) etc described in the wiki at https://trac.torproject.org/projects/tor/wiki/doc/weather-in-2014 Most of my work can be found at https://github.com/lucyd/tor-weather. Please feel free to provide feedback of any sort. Cheers, Sreenatha

1 0

Pluggable Transport Combiner Status Report 1
by quinn jarrell 07 Jun '14

07 Jun '14

Hi everyone, I'm the GSoC student working on the the Pluggable Transport Combiner. So here's what I've been working on: - I added in a configuration class and file reader to obfs-flashproxy client. - The obfs-flashproxy client can now launch any number of transports in a row. - It chains together the pluggable transports with SOCKS servers What I'll be working on next: - Documentation on the design - Making the combiner intercept the first tor connection instead of recieving the first pt's output as it does now - Testing authenticated transports. If you've got any questions or comments shoot me an email! Quinn Jarrell

2 1

GSoC Stegotorus update 1
by Noah Rahman 07 Jun '14

07 Jun '14

- Progress so far 1) Fixed segfaulting refactored SWF steganographic code. Code now fails on decompression but at least it's not segfaulting anymore. Suspected it was due to how trace payload server is integrated and rest of problems should be fixed by reimplementing pre decompression logic that was lost in refactored-away SWF_receive code. Can then move on to JS 2) Added config file for options using inih library, still testing locally. 3) With Yawning' s help, began to replace Moeller handshake with libelligator implementation. (#8089) Goals for next update 1) Finish SWF refactoring and associated unit test, start on JS 2) Finish elligator handshake and write unit test for config file. Test with ctgrind when I figure out how to patch valgrind properly. 3) Meet with Vinod at SRI and determine most expedient way of merging SRI changes into Stegotorus build (meeting scheduled for next Friday the 13th) Best Noah R.

1 0

Orbot & Orfox - Bi-weekly report
by Amogh Pradeep 06 Jun '14

06 Jun '14

Hey all, Attached to this email is my bi-weekly report on the project Orbot and Orfox.

1 0

[GSoC 2014] Revamp GetTor status report #1
by Israel Leiva 06 Jun '14

06 Jun '14

Hi all! This is the first status report for the Revamp GetTor GSoC project. I'll try to summarize what has been done. For more details feel free to ask. If you have comments, please tell me. Achievments: * Decided to rewrite the code. Made a proposal for a new GetTor's design. Available at [0]. * Discussed this new design with my mentors. Several changes has been made since its original version, including the use of ConfigParser for reading the links files. * Improved Python knowledge. Read PEP8 and PEP-0257 to (try to) create standard code. * Created the basis of the Core module, including a dummy script to test it. It can successfully read links from files and display them on screen. Available at [1]. Goal(s) until next report: * Have the Core module working with _at least_ one provider (Dropbox). This means adapt the current dropbox script to the new design. All of this properly documented and as standard as possible. That's all for now. Have a nice weekend! [0] https://github.com/ileiva/gettor/tree/master/spec [1] https://github.com/ileiva/gettor/tree/master/src -- israel

1 0

Get the latest TBB
by Israel Leiva 06 Jun '14

06 Jun '14

Hi all! Is there a better way to get a link for the latest TBB rather than looking at https://www.torproject.org/dist/torbrowser/? I'm thinking on something easy to automate (for the Revamp GetTor project). Thanks! -- israel

3 4

Proposal 236, Single-guard designs, and directory guards
by Nick Mathewson 06 Jun '14

06 Jun '14

Hi, all! I noticed that proposal 236 doesn't mention directory guards. (See proposal 207, implemented in Tor 0.2.4.) I think that we should consider retaining multiple directory guards while going to a single guard for multi-hop circuits. My rationale here is that when we have only a single directory guard, it can more easily perform hard-to-detect route biasing attacks by pretending not to have descriptors for nodes it doesn't like. Its ability to do this is limited by fact that we won't build circuits unless 95% of all paths are buildable (see get_frac_paths_needed_for_circuits() and its users). But still, trusting a single source for the completeness and freshness of your directory info is suboptimal. I also think that most of the arguments for single-guard apply to circuit guards more than to directory guards. But there could be some left, and we should figure those out. peace, -- Nick

5 6

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

tor-dev