[cabfpub] 答复: [EXTERNAL] 答复: Changing numbers of self-audited certificates

Thu Jun 8 07:01:08 MST 2017

Yes, I agree.

Let me explain it specifically, there is a premise when I proposed to set a minimum and a maximum value to certificate sample size during self-audit in my last email, and that is certificates should be audited separately according to different types(or different issuance processes) rather than selecting samples randomly from the overall certificates of a CA for audit. 

Best Regards

Ruby Xiong

Shanghai Electronic Certification Authority co., ltd. 

18F, No.1717, North Sichuan Road, Shanghai, China

Tel：+86-21-36393197

Email： <mailto:xiongyuanyuan at sheca.com> xiongyuanyuan at sheca.com 

发件人: Public [mailto:public-bounces at cabforum.org] 代表 Ryan Sleevi via Public
发送时间: Thursday, June 8, 2017 12:04 AM
收件人: CA/Browser Forum Public Discussion List
抄送: Ryan Sleevi
主题: Re: [cabfpub] [EXTERNAL] 答复: Changing numbers of self-audited certificates

>From the browser perspective, our target is 100% audit, particularly around technical controls. Recognizing the practical limitations of that, we've been willing to go less.

However, as a practical matter, the choice and application of the sampling, as currently practiced, does not align with the risk profile. For example, it would be entirely appropriate and reasonable to highlight that the issuance of certificates to different profiles (e.g. DV, OV, EV in one dimension, internally operated vs externally operated sub-CAs in another dimension, those using 3.2.2.4.1 vs 3.2.2.4.5, or those representing different CA brands or infrastructures) each represent a different set of activities and associated controls, and thus a more pragmatic approach to auditing would be to example sample sizes as appropriate for the populations of each of these distinct activities and controls, rather than the current approach that treats all certificate issuance as equivalent.

By taking a more meaningful examination of the practice - one in which every activity that exercises distinct controls is a distinct risk profile - it might be possible to reduce the sample size on an individual basis. Of course, in order to have reliable assurances of the meaningfulness of those samples, we would also need to see increased documentation in CA's CP/CPS (to highlight the activities) and in the audit reports (to detail the controls examined for each 'issuance pipeline')

On Wed, Jun 7, 2017 at 11:55 AM, Kirk Hall via Public <public at cabforum.org> wrote:

Hi, Ruby - thanks for the information.

If I recall correctly, we first required a 3% self-audit of certificates when the EV Guidelines were developed about ten years ago, but I can't remember the rationale, or why there was no maximum number of self-audits included.  Does anyone else remember?

-----Original Message-----
From: Public [mailto:public-bounces at cabforum.org] On Behalf Of xiongyuanyuan via Public
Sent: Wednesday, June 7, 2017 3:40 AM
To: 'CA/Browser Forum Public Discussion List' <public at cabforum.org>
Cc: xiongyuanyuan <xiongyuanyuan at sheca.com>
Subject: [EXTERNAL][cabfpub] 答复: Changing numbers of self-audited certificates

Yes.
>From the point of audit risk control, set a minimum value is more reasonable. According to the presentation in AICPA Audit Sampling Guide(screenshot attached), when a control happens less frequently, we can decide the sample size by the frequency of the control. So in my opinion, when CA performs self-audit to certificates that have a small volume, it is acceptable to take this guide as reference. From this table, we can see that a minimum value of 5 to certificate sample size is appropriate and is able to control audit risk.

Besides, I think we should also set a maximum value to certificate sample size.
This is because when CA performs self-audit to certificates that have a very large volume, 3% of the total population will still be a lot to audit, and this would result in large audit cost for the CA.
When we look at AICPA Audit Sampling Guide and AU 350 of PCAOB, for those testing samples with high risk, the TER（tolerable exception rate）should be low as possible, a maximum value of 60(certificates) to sample size(assume all the 60 testing samples are effective) will promise a lower TER as 5% which is much lower than a TER as12%-15% (which is operated by some audit firms for those testing samples with normal risk).

Base on this, I suggest we also set a maximum value of 60 to certificate sample size, which ensures audit efficiency as well as controls audit cost and audit risk.

Best Regards,
Ruby Xiong
Shanghai Electronic Certification Authority co., ltd.
18F, No.1717, North Sichuan Road, Shanghai, China
Tel：+86-21-36393197
Email：xiongyuanyuan at sheca.com

-----邮件原件-----
发件人: Public [mailto:public-bounces at cabforum.org] 代表 Gervase Markham via Public
发送时间: Tuesday, June 6, 2017 6:47 PM
收件人: CABFPub
抄送: Gervase Markham
主题: [cabfpub] Changing numbers of self-audited certificates

Currently, the BRs define, in section 8.7, the parameters for self-audits and audits of certificates below a TCSC. At the moment, the number of certs randomly chosen to be audited is defined as "the greater of one certificate or at least three percent of the Certificates issued".

I think that auditing just a single certificate (which is currently OK up until 33 are issued) makes it too easy to overlook problems when volumes are small. I propose instead a 5-certificate minimum, or 3%, whichever is larger. In other words:

Issued Audited
0      0
1      1
.....
5      5
6      5
.....
166    5
167    6
.....

We could just change the "one" to a "five" if people thought it was obvious that if you've issued less than five, you just audit all of them. Or we could expand the text a bit to explicitly describe that.

I would be interested in feedback on the impact of this change. It's been proposed for the Mozilla policy but as it's a BR stipulation I thought we should try here first.

Gerv
_______________________________________________
Public mailing list
Public at cabforum.org
https://cabforum.org/mailman/listinfo/public

_______________________________________________
Public mailing list
Public at cabforum.org
https://cabforum.org/mailman/listinfo/public
_______________________________________________
Public mailing list
Public at cabforum.org
https://cabforum.org/mailman/listinfo/public

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://cabforum.org/pipermail/public/attachments/20170608/211183dd/attachment-0001.html>