Re: New (?) authentication scheme for local communications

From: Stephan Neuhaus (neuhaus@cs.uni-sb.de)
Date: 03/20/03 

From: Stephan Neuhaus <neuhaus@cs.uni-sb.de>
Date: Thu, 20 Mar 2003 14:36:25 +0100

Marek Zawadzki wrote:
> Hello,
>
> I've implemented a siple and portable method of authenticating a client
> process which claims it is being run under certain username.
> I'd like to ask you about your opinion about it.
> [...]
> "How to make sure the client process is run by the user it claims to be
> (PF_UNIX sockets)":
> [Method deleted]
> I'll appreciate your opinions about the scheme presented here as well as
> about alternative approaches.

Wouldn't it be simpler for the server just to check /proc to see who is
on the other side of the connection (a la lsof) and then to see if the
effective user ID of that process is the one that it's supposed to be (a
la ps)? (Well, perhaps it would not be simpler... :-))

That would also protect from the following attack:

1. Eve wants to claim that she is in fact Alice. She is in cahoots with
Mark, who has compromised Alice's account on machine X.

2. Eve initiates the protocol on machine Y. Eve's process runs with
eve's uid as its effective uid.

3. Sam, the server, asks Eve to create a file in Alice's home directory.

4. Eve calls Mark on machine X, who has compromised Alice's account on
machine X and creates the file. (We assume that Alice's home is or can
be mounted read-write on machine X.)

5. Eve tells Sam to go ahead with the protocol.

6. Sam checks for the file and it is indeed there.

Granted, if Alice's account is compromised on machine X, then there is
not much security for her anyway, but it does violate your protocol
(because Sam will accept Eve's process as having Alice's euid).

Fun,

Stephan

Relevant Pages

  • Re: Question on Public Key Cryptography
    ... I don't know this protocol, but probably M is a random key and this is ... Dec), where pk_A is A's public key and sk_B is B's secret key. ... I'm guessing that the attack is something along the lines of: ... To discover the key inside C_AB, Eve ...
    (sci.crypt)
  • Re: Forward secrecy from two RNGs
    ... You want a protocol with the following two characteristics: ... > 1) Alice can read the plaintext, given her RNG and her two private keys. ... > 2) Eve can't read the plaintext, given all of Alice's previous RNG ...
    (sci.crypt)