[Cscwg-public] Invalidity Date
ianmcm at microsoft.com
Wed Sep 8 19:52:13 UTC 2021
Considering we would want the behaviors to consistent across the wide variety of OS versions in the market, we can say confidently that we expect CAs to continue to use the RevocationDate field as they do today. Any use of the Invalidity Date extension is not planned to be consumed.
From: Cscwg-public <cscwg-public-bounces at cabforum.org> On Behalf Of Ian McMillan via Cscwg-public
Sent: Friday, September 3, 2021 3:21 PM
To: rob at sectigo.com; cscwg-public at cabforum.org; bruce.morton at entrust.com; Corey.Bonnell at digicert.com
Subject: [EXTERNAL] Re: [Cscwg-public] Invalidity Date
I am looking into the current, past, and future behaviors of Windows now to get a definitive answer.
From: Cscwg-public <cscwg-public-bounces at cabforum.org<mailto:cscwg-public-bounces at cabforum.org>> On Behalf Of Rob Stradling via Cscwg-public
Sent: Thursday, September 2, 2021 6:14 AM
To: Bruce Morton <bruce.morton at entrust.com<mailto:bruce.morton at entrust.com>>; cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org>; Corey Bonnell <Corey.Bonnell at digicert.com<mailto:Corey.Bonnell at digicert.com>>
Subject: [EXTERNAL] Re: [Cscwg-public] Invalidity Date
[Resending; hopefully this message won't get lost in a moderation queue/blackhole this time]
> However, in all the documentation I've seen regarding Authenticode, it appears that the revocation date is the value that is checked by Windows and invalidityDate is seemingly not used.
That matches our experience.
On 2010-11-12, I received the following email from Tom Albertson, who at that time was in charge of the Microsoft Root Program:
I'm in over my technical head on this one, so treat it as more of a relay than anything else. When folks over here were looking at recent UserTrust CRLs, they noticed errors in Windows parsing the revocation date used. I'm not sure if it is a recent change or something you have been doing for a reason, but in any event:
In parsing CRLS, we populate the "Revocation Date" with the effective revocation date, but UserTrust is using the Invalidity Date extension in its CRLs. RFC 5280 defines the Invalidity Date extension as "a non-critical CRL entry extension that provides the date on which it is known or suspected that the private key was compromised or that the certificate otherwise became invalid." This extension has been around in the standards since 1999 at least as a recommended (SHOULD) extension. However, Windows has never supported it. Windows sets the effective revocation date in the RevocationDate field, which is supported by other code signing CAs. Or at least we haven't noted this use of the Invalidity Date extension by other CAs so far.
Can you look into this practice on your end, and try to find out the reason for it? Would there be any problem going forward indicating the effective revocation date in the RevocationDate field? This would appear to require re-issuing the CRLS, but not require rolling over any of your certificates.
Thanks and best regards,
The end result of that conversation was that we felt we had to treat Tom's request as a requirement from a root store operator that overrode RFC5280, which meant that we had to change our (previously RFC5280-compliant) implementation to start putting the effective revocation date into the "Revocation Date" field instead of the "Invalidity Date" extension. Since we haven't heard anything new from Microsoft on this topic since then, our implementation still behaves this way today. (I commented rather than deleted our original code, in the hope that we would one day be permitted to return to being RFC5280-compliant).
I don't like it when any aspect of policy is defined by a private communication that a root store operator sent only to a subset of CAs, but that seems to be what happened in this case.
> Could Ian or Mike confirm Windows's behavior in this regard?
An official, public update on Microsoft's policy requirements for encoding the effective revocation date in CRLs would also be much appreciated!
From: Cscwg-public <cscwg-public-bounces at cabforum.org<mailto:cscwg-public-bounces at cabforum.org>> on behalf of Corey Bonnell via Cscwg-public <cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org>>
Sent: 25 August 2021 22:50
To: Bruce Morton <bruce.morton at entrust.com<mailto:bruce.morton at entrust.com>>; cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org> <cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org>>
Subject: Re: [Cscwg-public] Invalidity Date
CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
I agree that using the invalidityDate CRL entry extension to express when the key corresponding to a revoked code signing certificate can no longer be trusted as opposed to the revocation date is conceptually cleaner and more in line with 5280 (which states that the revocation date SHOULD NOT be backdated such that it is before the issue date of the latest CRL).
However, in all the documentation I've seen regarding Authenticode, it appears that the revocation date is the value that is checked by Windows and invalidityDate is seemingly not used.
Could Ian or Mike confirm Windows's behavior in this regard?
From: Cscwg-public <cscwg-public-bounces at cabforum.org<mailto:cscwg-public-bounces at cabforum.org>> On Behalf Of Bruce Morton via Cscwg-public
Sent: Wednesday, August 25, 2021 1:59 PM
To: cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org>
Subject: [Cscwg-public] Invalidity Date
CSBR 13.2.1 states: A Certificate MAY have a one-to-one relationship or one-to-many relationship with the signed Code. Regardless, revocation of a Certificate may invalidate the Code Signatures on all signed Code, some of which could be perfectly sound. Because of this, the CA MAY specify a revocation date in a CRL or OCSP response to time-bind the set of software affected by the revocation, and software should continue to treat objects containing a timestamp dated before the revocation date as valid.
The CSBRs are referring to "revocation date', which I believe should be referring to "invalidity date" as specified in RFC 5280, https://datatracker.ietf.org/doc/html/rfc5280#section-5.3.2<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdatatracker.ietf.org%2Fdoc%2Fhtml%2Frfc5280%23section-5.3.2&data=04%7C01%7Cianmcm%40microsoft.com%7Cbfbdbc93dfc749d3988108d96f0ffa31%7C72f988bf86f141af91ab2d7cd011db47%7C0%7C0%7C637662936702792264%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=6cyXtkmkRmjnikatKAnmOqOIzwVG2OdeFWBXLCNfo2U%3D&reserved=0>.
Note that we need to think of the following dates:
* Valid from
* Invalidity date
* Revocation date
* Valid to
The purpose of the Invalidity date is to provide a date in the past, when the key was compromised. The revocation date would be on the date that the certificate was revoked and cannot be a past date.
Would there be any objections in changing "revocation date" to "invalidity date" in a future ballot?
Any email and files/attachments transmitted with it are confidential and are intended solely for the use of the individual or entity to whom they are addressed. If this message has been sent to you in error, you must not copy, distribute or disclose of the information it contains. Please notify Entrust immediately and delete the message from your system.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Cscwg-public