SEMI-OT: Prohibiting RC4 Cipher Suites

Viktor Dukhovni ietf-dane at dukhovni.org
Fri Feb 20 19:47:09 CET 2015


On Fri, Feb 20, 2015 at 07:26:00PM +0100, Patrick Ben Koetter wrote:

> A little off topic for DANE users, but somehow in scope. You might consider
> disabling RC4 in your servers cipher suite. IETF released an RFC requiring
> 
>    (...) that Transport Layer Security (TLS) clients and servers never
>    negotiate the use of RC4 cipher suites when they establish connections.
>    This applies to all TLS versions.  This document updates RFCs 5246, 4346,
>    and 2246.
>    -- Prohibiting RC4 Cipher Suites, https://tools.ietf.org/rfc/rfc7465.txt

However, this is not a good fit for opportunistic TLS in SMTP.

Leave RC4 enabled for now, there are still MTAs (mostly Microsoft
Exchange 2003) in the field that can only do RC4-SHA, RC4-MD5 or
a broken 3DES implementation.  They prefer RC4.  If you drop RC4,
you lose interoperability with these systems.

I would wait to disable RC4 for another 2-3 years.  With
Postfix one might do this instead:

    # Assumption, anyone with working 3DES can also do either AES
    # or RC4 with RC4 as the only remaining option rather rare.
    #
    smtpd_tls_exclude_ciphers = EXPORT, LOW, MD5, 3DES
    tls_preempt_cipherlist = yes

with these settings, the server chooses the best ciphersuite proposed
by the client based on its own preference order, not the client's.
By disabling no longer used EXPORT and <=64 bit (LOW) ciphers *and*
3DES, you still interoperate with any machines that only do RC4,
but it is only used as a *last* resort.

Fully disabling RC4 loses connectivity with some sites or best case
downgrades some email to cleartext instead, which is not more secure
than RC4.

Admittedly, if the server honours the clients cipherlist preferences
(default OpenSSL behaviour and also Postfix default tls_preempt_cipherlist
= no), then you might sometimes negotiate RC4 with clients that
can do AES, but chose to propose RC4 at a higher preference.

If you see evidence of this in your logs, you can apply the above
work-around, otherwise just leave things as they are.

-- 
	Viktor.


More information about the dane-users mailing list