[cabfpub] Pre-Ballot - Short-Life Certificates
jeremy.rowley at digicert.com
Wed Nov 5 11:56:11 MST 2014
I don't think their responses were encompassing as you indicated that all 9 points are cons. The comments received responded to points 7, 8, and 9 (plus Rick's general comments). Each are addressed below.
The response to 7 is that the revocation and expiration message aren't as strong. Gerv posted on the Mozilla list that they'd consider using the same message for expired short-lived certs as revoked certs. Therefore, I don't think this point has merit.
The response to 8 was that cert might need to be included in a CRL if the key is used for multiple certs. That's not really an argument against short-lived certs. Instead, that's an argument that users should have to frequently change private keys, which is more likely with a short-lived cert. I don't see that as a security issue for using short lived certs.
The rebuttal to point 9 was that CAs can use a shorter period. However, with short-lived certs the user gets to set the revocation period (up to 3 days) instead of the CA. This rebuttal missed the point that short-lived certs let users control the window of risk rather than delegated control to the CA.
Ricks points were:
"It would be helpful to list the disadvantages. Here's some:
a) Short-lived certs won't work on some number of machines whose clocks are too far out of sync. The actual number is debatable. It's small, but non-zero."
[JR] Gerv posted clock skew data. I think this data refutes this argument. Plus, does it matter? If a CA is offering short-lived certs, the CA takes the risk of clock skew. It doesn't affect any CA not issuing short-lived certs.
"b) Subscriber has to deal with the additional risk that their servers have to call back to the CA every day, obtain a fresh cert, and deploy that new cert without any negative impact to existing traffic."
[JR] Okay, but that's the subscriber's choice. If the subscriber warrants that the risks are worth it, why not let them use it? This is not a security argument against short-lived certs.
"c) What a CA saves in OCSP infrastructure is offset by new infrastructure needed to distribute new certs to these customers every day."
[JR] That's up to the CA. This is not an argument for allowing short-lived certs. Some CAs may not wish to offer short-lived certs, but that's no reason for the Forum to needlessly restrict an alternative form of revocation checking.
So far, the arguments against permitting short lived certs seems to be FUD. Are there any real reasons the Forum isn't permitting this type of cert?
From: i-barreira at izenpe.net [mailto:i-barreira at izenpe.net]
Sent: Wednesday, November 5, 2014 9:24 AM
To: Jeremy Rowley; public at cabforum.org
Subject: RE: [cabfpub] Pre-Ballot - Short-Life Certificates
Jeremy, my concerns were addressed later in emails by Eddy, Rick, etc.
Responsable del Área técnica
i-barreira at izenpe.net
ERNE! Baliteke mezu honen zatiren bat edo mezu osoa legez babestuta egotea. Mezua badu bere hartzailea. Okerreko helbidera heldu bada (helbidea gaizki idatzi, transmisioak huts egin) eman abisu igorleari, korreo honi erantzuna. KONTUZ!
ATENCION! Este mensaje contiene informacion privilegiada o confidencial a la que solo tiene derecho a acceder el destinatario. Si usted lo recibe por error le agradeceriamos que no hiciera uso de la informacion y que se pusiese en contacto con el remitente.
De: Jeremy.Rowley [mailto:jeremy.rowley at digicert.com]
Enviado el: miércoles, 05 de noviembre de 2014 16:19
Para: Barreira Iglesias, Iñigo; public at cabforum.org
Asunto: Re: [cabfpub] Pre-Ballot - Short-Life Certificates
Then please do so, and I will respond.
With a three day time period, I really don't see the cons. The cert will expire by the time any CA revokes. Nearly every CP states that they process revocation requests in 24 hours, meaning there is at least a day before it shows up on an OCSP and 10 before it propagates everywhere.
On 11/5/2014 3:19 AM, i-barreira at izenpe.net wrote:
> Jeremy, the 9 pros you mention can also be converted in cons. Don´t
> see it that easy as you indicate, in fact, I see more cons than pros
> Iñigo Barreira
> Responsable del Área técnica
> i-barreira at izenpe.net
> ERNE! Baliteke mezu honen zatiren bat edo mezu osoa legez babestuta egotea. Mezua badu bere hartzailea. Okerreko helbidera heldu bada (helbidea gaizki idatzi, transmisioak huts egin) eman abisu igorleari, korreo honi erantzuna. KONTUZ!
> ATENCION! Este mensaje contiene informacion privilegiada o confidencial a la que solo tiene derecho a acceder el destinatario. Si usted lo recibe por error le agradeceriamos que no hiciera uso de la informacion y que se pusiese en contacto con el remitente.
> -----Mensaje original-----
> De: public-bounces at cabforum.org [mailto:public-bounces at cabforum.org]
> En nombre de Jeremy Rowley Enviado el: viernes, 31 de octubre de 2014
> Para: public at cabforum.org
> Asunto: Re: [cabfpub] Pre-Ballot - Short-Life Certificates
> I think this discussion could benefit from some of the reasons that people want short-lived certificates. These were previously discussed on the Mozilla mailing list, but not everyone follows what is going on there.
> Here are some of the advantages:
> 1) Subscribers in areas prone to unrest and where their server might be taken over can let the certificate expire, essentially letting the certificate fail to renew.
> 2) Subscribers can eliminate size from the certificate
> 3) Subscribers can avoid call-backs to the CA
> 4) Subscribers can control how long revocation information takes to
> propagate (up to 3 days, but could be less)
> 5) This is a change that doesn't require action by the browsers,
> meaning it's a CA-drive improvement
> 6) The change isn't required for any CA. All it does is permit CAs interested in offering the services to customers to do so.
> 7) Short-lived certs provide a limited hard-fail since the expiration
> message for expired certs is more visible than the message received
> where revocation information is unavailable
> 8) Browsers don't need to add the certs to their CRLSets or do a call to the CA to retrieve revocation information.
> 9) Short-lived certs provide shorter revocation windows than currently offered under the BRs.
> There are 9 advantages that I could readily find. I'm sure many more exist. Of course, short-lived certs aren't for every customer and will be deployed only by a small percent of the internet population. However, for those interested in using them, there are key advantages in deployment.
> -----Original Message-----
> From: public-bounces at cabforum.org [mailto:public-bounces at cabforum.org]
> On Behalf Of Gervase Markham
> Sent: Thursday, October 30, 2014 7:29 AM
> To: kirk_hall at trendmicro.com; Ryan Sleevi; Doug Beattie
> Cc: public at cabforum.org
> Subject: Re: [cabfpub] Pre-Ballot - Short-Life Certificates
> On 29/10/14 18:50, kirk_hall at trendmicro.com wrote:
>> Ryan, thanks for the information, and I respect your analysis. But
>> many of us would say that revocation (and the ability to check for
>> revocation) is a fundamental aspect of whether a cert is valid at all.
> I think we all agree that the ability to revoke certs is vital. However, in the real world, there is always going to be a time lag of some sort between the decision to revoke and all clients becoming aware of that revocation. Comparing any system to the perfect system of universal instant revocation is unfair.
> An analysis of real-world revocation inevitably involves complex scenarios about the nature of the attack, the capabilities of the attackers, the type of revocation system being used, the update frequency of clients, the characteristics of the network, and so on. And it inevitably involves vulnerability windows for some clients.
> My assertion is that, in reasonable attack scenarios, the vulnerability windows and overall risk of short-lived certs is about the same as long-lived certs using OCSP.
> Public mailing list
> Public at cabforum.org
> Public mailing list
> Public at cabforum.org
More information about the Public