Skip to content

Instantly share code, notes, and snippets.

@chelseakomlo

chelseakomlo/existing_tor_guard_selection.md

Last active February 11, 2016 15:48
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save chelseakomlo/2acbe15314b5a809c6f4 to your computer and use it in GitHub Desktop.
Save chelseakomlo/2acbe15314b5a809c6f4 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
existing_tor_guard_selection.md

Existing Tor Guard Selection Algorithm

  • ALL_GUARD_LIST = guard information from latest consensus
  • GUARD_LIST = guards persisted to our state file
  • DIRECTORY_GUARD = if we select guards with the V2Dir flag. Guards with the V2Dir Flag can be used as entry guards for both fetching information from directories as well as for standard entry guards.

ON_BOOTSTRAP (no existing guards)

  1. RECEIVE_NEW_CONSENSUS
  2. From listed guards in ALL_GUARD_LIST with DIRECTORY_GUARD=true: 1. 3 times do (default guard value on startup):
    1. ADD_RANDOM_ENTRYGUARD to choose a guard
    2. Add this new guard to GUARD_LIST

RECEIVE_NEW_CONSENSUS

  1. Mark guards that are not listed in the latest consensus as "bad" in ALL_GUARD_LIST
  2. Remove guards that have been dead for 30 days from GUARD_LIST and ALL_GUARD_LIST
  3. Remove guards that were added more than 30 days ago from GUARD_LIST and ALL_GUARD_LIST

BUILD_NEW_CIRCUIT

  1. First, CHOOSE_A_GUARD
  2. Then CONNECT_ENTRY_GUARD with our chosen guard

CHOOSE_A_GUARD

  1. Ensure that we have enough entry guards (only need 1) 1. If we do not:
    1. Use the ADD_RANDOM_ENTRYGUARD algorithm to choose a new guard, DIRECTORY_GUARD=false
    2. Add new guard to GUARD_LIST
  2. From GUARD_LIST of entry guards: 1. Build list of live guards, that are:
    1. listed in the current consensus, except if:
    2. It is offline and and we have previously tried 2. From this list of live guards
    3. Choose the number of guards that we need if the list contains:
    4. Guards that we have not tried, or
    5. Or if the list contains at least the number that we needed (which is 1) 3. Otherwise (if the list of live guards contains only 1):
    6. Relax our constraints (bandwidth, uptime, for directory, etc), and
    7. Use the ADD_RANDOM_ENTRYGUARD algorithm to choose a new guard
    8. Try to choose a new guard, starting over from b.

ADD_RANDOM_ENTRYGUARD

  1. Build weighted distribution of all guards based on bandwidth
  2. Pick a guard at random from this distribution

CONNECT_ENTRY_GUARD

  1. If we have never made contact with this guard before: 1. If we can connect:
    1. Mark guard as we have made contact
    2. Assume the network was down, and mark all guards in GUARD_LIST for retry (except the guard that we just connected) 2. If we cannot connect:
    3. Remove it from GUARD_LIST
    4. [what is the fallback behavior? Do we choose another guard from GUARD_LIST]?
  2. If we have made contact with this guard before: 1. If we cannot connect:
    1. Mark this guard as offline
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment