#32 don't auto-block remailers address

closed-fixed
Remailer (20)
8
2002-12-14
2002-10-11
Trek Star
No

mixmaster auto-block feature now check (in
rem.c:blockrequest()) if the address to be blocked is
not the remailer address.
It should check also if the address is not of a
remailer or ignore it, because an attacker can send a
spoofed request to block a remailer from the remailer
network.

Also before to send reply.txt (that contains a
DESTINATION-BLOCK line) it should check if the
destination address is a remailer.

Discussion

  • Len Sassaman
    Len Sassaman
    2002-10-18

    • priority: 5 --> 8
    • assigned_to: nobody --> weaselp
     
  • Len Sassaman
    Len Sassaman
    2002-10-18

    Logged In: YES
    user_id=29569

    Well, now it's public. How do you want to handle this? If we
    simply ignore remailer entries in the .blk file, then remops
    won't be able to block mail to other remailers (a common
    practice when trying to minimize the damage donw by a dead
    remailer.)

    If we do the check against known remailers when the
    destination-block requests come in, that would seem better.

     
  • Len Sassaman
    Len Sassaman
    2002-10-21

    Logged In: YES
    user_id=29569

    Okay, here's how the fix should work. Before we check if an
    address is in dest.blk, we should check if it is an address in
    our keyring(s). If it is, we should ignore dest.blk and send the
    message.

    (This blocks the attack, and still allows remops to block dead
    remailers -- they just have to remove the dead remailer's key
    from their key rign first.)

    One might also argue that middleman remailers should
    perform a check of their keyrings before sending any mail,
    also. (Peter says that middleman remailers allow or deny
    messages to be sent based on message type, which opens
    up certain mailbombing attacks.) We could fix both of these
    in a similar fashion, though I am primarily concerned about
    the first one.

     
  • Logged In: YES
    user_id=1525

    fixed in CVS

     
    • status: open --> open-fixed
     
  • Len Sassaman
    Len Sassaman
    2002-12-14

    • status: open-fixed --> closed-fixed