[tor-bugs] #7144 [Core Tor/Tor]: Implement Bridge Guards and other anti-enumeration defenses

[Tor Bug Tracker & Wiki]

#7144: Implement Bridge Guards and other anti-enumeration defenses
-------------------------------------------------+-------------------------
 Reporter:  karsten                              |          Owner:  isis
     Type:  project                              |         Status:
                                                 |  needs_revision
 Priority:  High                                 |      Milestone:  Tor:
                                                 |  unspecified
Component:  Core Tor/Tor                         |        Version:
 Severity:  Normal                               |     Resolution:
 Keywords:  SponsorZ, tor-bridge,                |  Actual Points:
  027-triaged-1-out, 028-triage, 028-triaged,    |
  isis201604, isis201605, TorCoreTeam-           |
  postponed-201604, nickm-deferred-20160905,     |
  tor-03-unspecified-201612                      |
Parent ID:                                       |         Points:  3
 Reviewer:                                       |        Sponsor:
                                                 |  SponsorS-can
-------------------------------------------------+-------------------------

Comment (by larsl):

 A question about bridge guards in general: can't they be used by the next
 hop after the guard to passively verify that the bridge guard really is a
 bridge guard? Consider a client C and a bridge B with a bridge guard G. C
 doesn't know that G is a bridge guard of B, so at some point it tries to
 open the circuit C -> B -> R -> G where R is some arbitrary relay. B
 inserts a loose-source routed hop to G, since that's its bridge guard, so
 we get

    C -> B -> G -> R -> G

 Since no one will ever try to create a circuit like that directly, since R
 will refuse to extend through G, R now knows that the first hop through G
 must be loose-source routed, and thus G must be a bridge guard. Neither C
 nor B can prevent this since C doesn't know which guards B are using, and
 B can't read the encrypted extend cells intended for R.

 Is this bad, or do we assume that bridge guards are detectable anyway? Or
 is it actually prevented by the path-choosing algorithms somehow?

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/7144#comment:46>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online