Re: [TLS] Is it possible for a client to offer TLS 1.3, but not be forced to support RSA PSS in TLS 1.2?

[Andrey Jivsov <[e-mail]>]

On 05/29/2018 12:42 PM, Benjamin Kaduk wrote:
> On Tue, May 29, 2018 at 12:35:20PM -0700, Andrey Jivsov wrote:
>> On 05/29/2018 12:13 PM, Benjamin Kaduk wrote:
>>> On Tue, May 29, 2018 at 11:57:39AM -0700, Andrey Jivsov wrote:
>>>> Greetings.
>>>>
>>>> TLS 1.3 draft in sec 4.2.3.  Signature Algorithms tells that if a client
>>>> wants to negotiate TLS 1.3, it must support an upgraded (and
>>>> incompatible) version of TLS 1.2, the one that changes RFC 5246 to allow
>>>> RSA-PSS in sec. 7.4.1.4.1. Signature Algorithms.
>>>>
>>>> You might recall that the possibility to negotiate between PSS and
>>>> RSASSA-PKCS1-v1_5 in TLS 1.3 handshake, just as it is allowed for X.509
>>>> signatures, was discussed on the mailing list. The WG decision then was
>>>> to hard-wire PSS in the TLS 1.3 handshake.
>>>>
>>>> I don't recall any discussion on going further than this, all the way to
>>>> changing the 10-year old TLS 1.2.
>>>>
>>>> Unfortunately, our products have issues with PSS beyond our control. The
>>>> only solution left to avoid receiving PSS with TLS 1.2 is to never
>>>> negotiate TLS 1.3 as a client. Another solution is insecure fallback,
>>>> but we presently don't do this.
>>>>
>>>> Is my reading of the situation correct? Thank you.
>>>
>>> Sounds like it:
>>>
>>>    RSASSA-PKCS1-v1_5 algorithms  Indicates a signature algorithm using
>>>       RSASSA-PKCS1-v1_5 [RFC8017] with the corresponding hash algorithm
>>>       as defined in [SHS].  These values refer solely to signatures
>>>       which appear in certificates (see Section 4.4.2.2) and are not
>>>       defined for use in signed TLS handshake messages, although they
>>>       MAY appear in "signature_algorithms" and
>>>       "signature_algorithms_cert" for backward compatibility with TLS
>>>       1.2,
>>>
>>> -Ben
>>>
>>
>> I was referring to
>>>
>>>    -  Implementations that advertise support for RSASSA-PSS (which is
>>>       mandatory in TLS 1.3), MUST be prepared to accept a signature
>>>       using that scheme even when TLS 1.2 is negotiated.  In TLS 1.2,
>>>       RSASSA-PSS is used with RSA cipher suites.
>>
>> I am OK with what you quoted. What I just quoted represents a
>> significant change in behavior in TLS 1.2 and there is no way to opt out
>> of this change to TLS 1.2.
> 
> Ah, I misread your original message, but all is clear now.
> 
>> I will add that I've seen this behavior by servers already, even when
>> client doesn't advertise TLS 1.3. Just the fact of including some 08 xx
>> IDs in signature_algorithms in ClientHello, without protocol_version
>> extension, gets the TLS 1.2 upgraded to RSA-PSS.
>>
>> IMO this paragraph should be removed. Those that want PSS in the
>> handshake should negotiate TLS 1.3. Preservation of current behavior of
>> TLS 1.2 is important, at least as an option.
> 
> First off, it's basically too late to make substantive changes like that;
> the bar to meet is something like "a huge outcry from deployments" or
> "a critical security flaw".
> 
> Second, what's going on here is that TLS 1.3 is defining some new signature
> algorithms for TLS messages, and making them mandatory to support for TLS 1.3.
> But negotiation of TLS signature algorithms has *always* been independent of
> protocol version.  If you support TLS 1.3, you also support the new signature
> algorithms; if you support TLS 1.3 and TLS 1.2, you support the new signature
> algorithms and you support TLS 1.2, therefore by the longstanding negotiation
> rules you are obligated to support the combination.  You are in effect proposing
> that we make a break in the signature (and hash) algorithm space with individual
> algorithms supported either in <=1.2 or >=1.3, but not both -- we did this for
> ciphersuites since we fundamentally changed the meaning of what a ciphersuite is.
> But the signature scheme does not seem to have undergone such a fundamental change,
> so it seems hard to justify introducing this sort of split.
> 
> -Ben
> 

We are talking about TLS 1.3-specific IDs:

         /* RSASSA-PSS algorithms with public key OID rsaEncryption */
          rsa_pss_rsae_sha256(0x0804),
          rsa_pss_rsae_sha384(0x0805),
          rsa_pss_rsae_sha512(0x0806)

In TLS 1.2 08 corresponds to the (undefined) hash algorithm, and thus
these IDs have no meaning to TLS 1.2.

TLS 1.3 spec purposely "split" these IDs so that they have no meaning to
TLS 1.2 servers. One needs the paragraph I quoted to force
implementations to change the behavior of TLS 1.2.

What's the harm of dropping this one paragraph I quoted and keeping TLS
1.2 behavior the same?