[tor-bugs] #4817 [Tor]: Control port authentication failures don't differentiate failure types

[Tor Bug Tracker & Wiki]

#4817: Control port authentication failures don't differentiate failure types
-------------------------+-------------------------------------------
     Reporter:  atagar   |      Owner:
         Type:  defect   |     Status:  needs_review
     Priority:  trivial  |  Milestone:  Tor: 0.2.5.x-final
    Component:  Tor      |    Version:
   Resolution:           |   Keywords:  easy maybe-proposal tor-relay
Actual Points:           |  Parent ID:
       Points:           |
-------------------------+-------------------------------------------

Comment (by nickm):

 So, (B) and (C) are not technically distinct: You are allowed to specify a
 password in hex, and passwords are allowed to be 32 bytes long.  So you
 can also say: ```AUTHENTICATE 6d792070617373776f7264` instead of
 ```AUTHENTICATE "my password"```.

 So if a controller says ```AUTHENTICATE
 b4c9e2effc93bbbf139dcc5c0fc15d0b890a9e7bf7c8bb49b1d34c2eb547c910```, we
 don't actually ''know'' that they're providing a cookie rather than
 providing a very strange password and encoding it in hex.

 Similarly, (E) and (F) are not totally distinct: passwords are allowed to
 be 32 characters, and users are allowed to send control-cookies as
 C-encoded strings if they choose.

 In other words, we can't actually determine which kind of authentication
 the controller was trying to send.  We know that non-32-byte fields are
 never cookies, and that's about it.  We can *guess* that 32-byte hex-
 encoded things are ''usually'' cookies, but that's only a heuristic.

 Is this still worth doing IYO?

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/4817#comment:11>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online