Re: [dhcwg] on use case discussion for draft-ietf-dhc-sedhcpv6

[神明達哉 <[e-mail]>]

At Wed, 04 Mar 2015 21:36:24 +0100,
Francis Dupont <[e-mail]> wrote:

> >  It's still vague to me...specifically what kind of attack should we
> >  worry about in the case of client->server authentication when we use
> >  the shared secret mechanism, and in which sense is it more serious
> >  than with the public/private key mechanism?
>
> => (forgot this one) in the client->server case with shared secrets
> you can attack the client or the server (and for the server get all
> the secrets), with public/private key pairs only an attack against
> the client can reveal the secret/private key.

True, but if the purpose of the authentication is to determine whether
the DHCP server should provide specific clients with DHCP server
(assigning an address, etc), compromising the server can defeat it
anyway; the attacker doesn't have to know clients' private keys.

> So public/private key mechanisms are intrinsically more secure than
> shared secret mechanisms.

As a general principle of public key-based vs shared secret-based,
yes.  But the important part of my question is "in the case of
client->server authentication".

> BTW public/private key mechanisms have some operational advantages
> too because they don't require a way to communicate the shared secret
> between the two parties before use.

Right, and if it's used in cases where TOFU is effective, I already
agreed on that.  But in cases where public keys needed to be shared
securely, I don't see the clear operational advantages for public
key-based mechanisms.

--
JINMEI, Tatuya