[Cscwg-public] Timestamp Certificate and SubCA updates

Ian McMillan ianmcm at microsoft.com
Thu Apr 4 17:54:33 UTC 2024 

Hi Martijn and all,

Thinking more about this requirement and the way folks are operating today and I can see with this requirement CA may be constrained when they encounter a need for greater agility to scale out. I'd like to propose a means to keep issuing CAs for time stamping end-entity certificates online when they are shorter-lived certificates (less than 72 months in validity). I agree that we need wording to say existing CAs are not required to adhere to these clarified requirements, but all new subCAs must meet the requirements. Here is my proposal...

"'Effective April 15, 2025, a Timestamp Authority MUST protect all Private Keys associated with its Root CA certificates and new Subordinate CA certificates with a validity period of greater than 72 months containing the `id-kp-timeStamping` KeyPurposeId in the `extKeyUsage` extension (per section 7.1.2.2 g) and that issued Timestamp Certificates with the policyidentifier 2.23.140.1.4.2, in a Hardware Crypto Module conforming to the requirements specified in [Section 6.2.7.1](#6271-private-key-storage-for-CA-keys), maintained in a High Security Zone and in an offline state or air-gapped from all other networks.'

What my intent for this to mean is that all new subCAs issuing time stamp certificates will fall into two buckets....


  *   SubCA Certificate validity greater than 72 months MUST/SHALL be secured offline
  *   SubCA Certificate validity less than 72 months MAY be secured online (or 'SHOULD be secured offline')

This is similar to a policy we've operated under internally with the risk of online issuing CAs where we need greater agility to scale out as the business needs for certificate issuance/fulfillment have higher volumes and latency requirements that offline CAs just cannot effectively meet.

Thanks,
Ian

From: Cscwg-public <cscwg-public-bounces at cabforum.org> On Behalf Of Martijn Katerbarg via Cscwg-public
Sent: Thursday, April 4, 2024 6:07 AM
To: Mohit Kumar <mohit.kumar at globalsign.com>; cscwg-public at cabforum.org
Subject: [EXTERNAL] Re: [Cscwg-public] Timestamp Certificate and SubCA updates

Hi Mohit,

> Can I confirm that the proposal to protect private keys of Subordinate CAs in an offline state is applicable to only private keys generated for Roots/Subordinate CAs created after the effective date.

You're touching on a good point here. The way the requirement is written, can be interpreted as such that the CA would need to do this for existing SubCAs as well. I'm leaving out Root CAs here, since these even now, already are required to be in an offline state.

But, making this a requirement for existing SubCAs may be an issue... while CAs may be able to migrate keys to an offline HSM, that's not always a given. If that's not an option, then key destruction would also not be an option, cause even if the CA wouldn't use the SubCA is issue new timestamp certificates anymore, they would still need use the key for signing CRLs.

I think it makes sense to make the requirement to be for any SubCA still used for the issuance of timestamp certificates going forward. If there's no objection to that approach, I can update the ballot.

(I won't be able to make todays call, but please feel free to discuss and let me know the outcome)

> Also, per my understanding, the scope of the proposal is limited to only Root/Subordinate CAs issuing Timestamp Certificates for Code Signing (i.e. with the OID 2.23.140.1.4.2). If yes, may be it would be better to clarify the same with the following language update

Correct. I would expect Section 1.2 already makes clear that only TS certificates with this OID, assert compliance with the CSBRs.

As far from a root store standpoint, it seems the the MS Root store policy requires the Policy OID to be included for TLS and Non-EV code signing, but there is no mentioning there about Timestamping certificates specifically.

Regards,

Martijn

From: Mohit Kumar <mohit.kumar at globalsign.com<mailto:mohit.kumar at globalsign.com>>
Date: Thursday, 4 April 2024 at 04:19
To: Martijn Katerbarg <martijn.katerbarg at sectigo.com<mailto:martijn.katerbarg at sectigo.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] Timestamp Certificate and SubCA updates

Hi Martijn,



Can I confirm that the proposal to protect private keys of Subordinate CAs in an offline state is applicable to only private keys generated for Roots/Subordinate CAs created after the effective date.



Also, per my understanding, the scope of the proposal is limited to only Root/Subordinate CAs issuing Timestamp Certificates for Code Signing (i.e. with the OID 2.23.140.1.4.2). If yes, may be it would be better to clarify the same with the following language update



'Effective April 15, 2025, a Timestamp Authority MUST protect Private Keys associated with its Root CA certificates and Subordinate CA certificates containing the `id-kp-timeStamping` KeyPurposeId in the `extKeyUsage` extension (per section 7.1.2.2 g) and that issued Timestamp Certificates with the policyidentifier 2.23.140.1.4.2, in a Hardware Crypto Module conforming to the requirements specified in [Section 6.2.7.1](#6271-private-key-storage-for-CA-keys), maintained in a High Security Zone and in an offline state or air-gapped from all other networks.'



Thanks

Mohit


From: Cscwg-public <cscwg-public-bounces at cabforum.org<mailto:cscwg-public-bounces at cabforum.org>> On Behalf Of Martijn Katerbarg via Cscwg-public
Sent: Tuesday, March 19, 2024 5:04 AM
To: cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org>; Dimitris Zacharopoulos (HARICA) <dzacharo at harica.gr<mailto:dzacharo at harica.gr>>
Subject: Re: [Cscwg-public] Timestamp Certificate and SubCA updates

The language (https://github.com/cabforum/code-signing/pull/34 ) has been further updated (https://github.com/cabforum/code-signing/pull/34/commits/9288f7ec376b4bbd139dcb596bcb2d1bf9bd7683) based on the below. @Dimitris Zacharopoulos (HARICA)<mailto:dzacharo at harica.gr> I replaced "deleted" with "destroyed" in your proposal, as I believe it would fit better in that section.

Are there any further comments? If not I will start the official discussion period in the next few days.

Regards,

Martijn

From: Cscwg-public <cscwg-public-bounces at cabforum.org<mailto:cscwg-public-bounces at cabforum.org>> on behalf of Martijn Katerbarg via Cscwg-public <cscwg-public at cabforum.org<mailto:cscwg-public at cabforum.org>>
Date: Monday, 11 March 2024 at 09:51
To: Dimitris Zacharopoulos (HARICA) <dzacharo at harica.gr<mailto:dzacharo at harica.gr>>, 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] Timestamp Certificate and SubCA updates
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.

Works for me on both fronts. I'll leave the discussion open for a bit so others can add on.

From: Dimitris Zacharopoulos (HARICA) <dzacharo at harica.gr<mailto:dzacharo at harica.gr>>
Date: Monday, 11 March 2024 at 09:48
To: Martijn Katerbarg <martijn.katerbarg at sectigo.com<mailto:martijn.katerbarg at sectigo.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] Timestamp Certificate and SubCA updates
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.


On 11/3/2024 10:32 π.μ., Martijn Katerbarg wrote:
Thanks Dimitris, I've reviewed and accepted the suggestions.

> witnessed by members of two different Trusted Roles (not by two Trusted Role Members, i.e. you can't use two persons of the same Trusted Role).

TBH, I'm not sure why it couldn't be two persons of the same Trusted Role?

I'm not a native English speaker but I think "Roles" (plural) points to the different types of Roles, while "Trusted Role members" would point to different Members in any Trusted Role. If the intent is to have a 4-eye principle control from any Trusted Role, we can make it clearer by using the "Trusted Role members" phrase.

> In general, a "key destruction" ceremony includes the deletion of all copies of the key, including copies that reside in backups. If we require a "key destruction" ceremony, the "restore key" case is nonsensical. We probably need to work on this some more so that we all have the same understanding and expectations.
> It's ok to keep the keys in backups but if you happen to restore them in an HSM, you must not use them to sign anything. If a CA/TSA can also "destroy" the key, meaning that all copies of that private key can be unequivocally/securely deleted (i.e. without a way to recover the key), including any instance of the key as part of a backup, the better!
Agreed in general regarding the Key Destruction ceremony. However having to also destroy the backup of the key, and do this again for any next key every 18 months, can be a lengthy procedure, specially if backups are stored securely and offline in different places around the world. That's why for this case we specifically call out that backups don't need to be destroyed.

But your point on an HSM restoring an entire partition and that violating the requirement, is valid.

For reference, the current proposed language is:
The CA MAY maintain existing backup sets containing the Private Key corresponding to a Timestamp Certificate. The CA MUST NOT restore the Private Key corresponding to a Timestamp Certificate contained within the backup if the Timestamp Certificate was issued more than 15 months prior to restoration of the backup.

What about we once more use the NSR language and state:
The CA MAY maintain existing backup sets containing the Private Key corresponding to a Timestamp Certificate. The CA SHOULD NOT restore the Private Key corresponding to a Timestamp Certificate contained within the backup if the Timestamp Certificate was issued more than 15 months prior to restoration of the backup. If the CA does restore such a Private Key, the CA SHALL only restore the Private Key in a suitable HSM while it's maintained in a High Security Zone and in an offline state or air-gapped from all other networks and perform a new key destruction ceremony prior to the HSM being brought to an online state.

Thoughts?

If we want to allow the existence of a key in a backup, IMHO we should avoid using the "key destruction" language. How about the following:

Modify
For Timestamp Certificates issued on or after June 1, 2024, the CA SHALL log the removal of the Private Key from the Hardware Crypto Module through means of a key destruction ceremony performed by the CA and witnessed and signed-off by at least two Trusted Roles.

to
For Timestamp Certificates issued on or after June 1, 2024, the CA SHALL log the removal of the Private Key from the Hardware Crypto Module through means of a key deletion ceremony performed by the CA and witnessed and signed-off by at least two Trusted Role members. The CA MAY also perform a key destruction ceremony, meaning that all copies of that private key are unequivocally/securely deleted (i.e. without a way to recover the key), including any instance of the key as part of a backup, to satisfy this requirement.

Thanks,
Dimitris.

As a side-note, I wonder if there's a task for the NSWG (or Definitions WG once it's setup) to define terms for online and offline HSMs



From: Cscwg-public <cscwg-public-bounces at cabforum.org><mailto:cscwg-public-bounces at cabforum.org> on behalf of Dimitris Zacharopoulos (HARICA) via Cscwg-public <cscwg-public at cabforum.org><mailto:cscwg-public at cabforum.org>
Date: Sunday, 10 March 2024 at 10:30
To: 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] Timestamp Certificate and SubCA updates
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.

Hi Martijn,

Two suggestions submitted on GitHub.

Regarding the prohibition of restoring a private key of a Timestamp Certificate, I'm not sure how universal this can be because some HSMs restore an entire slot/partition, which might contain Private Keys associated with obsolete Timestamp Certificates. As the ballot is written, such an action would be a violation.

In general, a "key destruction" ceremony includes the deletion of all copies of the key, including copies that reside in backups. If we require a "key destruction" ceremony, the "restore key" case is nonsensical. We probably need to work on this some more so that we all have the same understanding and expectations.

Let me restate the intent of this requirement as discussed all this time, and please correct me if I'm wrong.

IMO, the goal is to put the keys associated with Timestamp Certificates out of use, 15 months after the notBefore of the Timestamp Certificate.

In order to achieve some level of assurance for this action, the proposal is to delete the keys from the HSM 18 months after the notBefore of the Timestamp Certificate, in an audited way, witnessed by members of two different Trusted Roles (not by two Trusted Role Members, i.e. you can't use two persons of the same Trusted Role).

It's ok to keep the keys in backups but if you happen to restore them in an HSM, you must not use them to sign anything. If a CA/TSA can also "destroy" the key, meaning that all copies of that private key can be unequivocally/securely deleted (i.e. without a way to recover the key), including any instance of the key as part of a backup, the better!

Thoughts?

Dimitris.
On 6/3/2024 2:07 μ.μ., Martijn Katerbarg via Cscwg-public wrote:
All,

As discussed last week, I'd send out the draft language for this ballot once more before starting the discussion period.  The latest version can be found in https://github.com/cabforum/code-signing/pull/34

I've made changes this morning to add 3 effective dates, these are:


  1.  For the removal of private keys associated with timestamp certificates, effective June 1st, 2024, CAs will need to properly log the removal of said key.

     *   While I expect CAs to already properly log this for audit purposes even now, there may be exceptions for when this has not been done,  for example a private key or timestamp certificate that was signed maybe 20 years ago. This language is added to avoid any confusion on from what point there needs to be an audit trail

  1.  Effective April 15, 2025, private keys associated with SubCAs containing the "Time Stamping" EKU will need to be placed in offline HSMs.

     *   I believe a roughly one year effective date is appropriate here, since CAs may need to move keys from one HSM to another.

  1.  For private keys associated with timestamp certificates that were issued for greater than 15 months, CAs will need to remove the private keys 18 months after certificate issuance, starting April 15, 2025.

     *   Likewise, I feel like anything involving HSM process changes, should have a longer effective date, and it makes sense to align this with the effective date above.

I'll start a ballot on this early next week, unless there is concern with the above.

Regards,

Martijn


_______________________________________________

Cscwg-public mailing list

Cscwg-public at cabforum.org<mailto:Cscwg-public at cabforum.org>

https://lists.cabforum.org/mailman/listinfo/cscwg-public


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cabforum.org/pipermail/cscwg-public/attachments/20240404/1b95e383/attachment-0001.html>

More information about the Cscwg-public mailing list