[Servercert-wg] Proposal to update logging requirements
martijn.katerbarg at sectigo.com
Wed Sep 20 09:04:29 UTC 2023
Thank you for your comments. I agree. With that, my existing proposal hopefully is a first step towards this.
I’m also contemplating if, long term, all logging requirements should instead be included in the NSRs, rather than the BRs. But, as the NSWG is already working on a set of changes, I can imagine this is not something that will (or should) happen soon.
From: Daniel Jeffery <djeffery at fastly.com>
Date: Friday, 15 September 2023 at 00:38
To: Martijn Katerbarg <martijn.katerbarg at sectigo.com>, CA/B Forum Server Certificate WG Public Discussion List <servercert-wg at cabforum.org>
Subject: Re: [Servercert-wg] Proposal to update logging requirements
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.
Hello Martijn and CA/B,
I like where we're going with this and wholeheartedly agree with the desire to not obscure useful logging with excessive volume of useless logs.
In that vein, I'm curious what uses people have for logging all blocked traffic on an internet facing firewall. To me it seems the signal to noise ratio is so bad that keeping all the logs of dropped packets on an external interface is unproductive. The only times I can really see using this is with some highly-tuned NIDS or in a retrospective to look at patterns prior to breach.
Logging all blocked firewall traffic behind the firewall, between security zones, on the other hand, should be very useful.
On Wed, 13 Sept 2023 at 03:00, Martijn Katerbarg via Servercert-wg <servercert-wg at cabforum.org <mailto:servercert-wg at cabforum.org>> wrote:
During our last WebTrust audit cycle it became clear that our interpretation of “Firewall and router activities” and CPA Canada’s interpretation were meaningfully different. In particular it came to light that in its most aggressive possible interpretation, the actual logging of a firewall activity would itself constitute a firewall activity, which would itself require logging, as would the log of the log entry of that log entry, the log of this newest log entry, and etcetera into infinity. In our opinion, too much “valid traffic” logging, makes it harder to find “bad traffic”.
We offer a simple rewrite to reflect the difference between valuable and necessary logged information and unproductive (and potentially absurd) logging.
Similarly, several Certificate Consumers have expressed the wish to move away from OCSP, while, depending on interpretation of the language, CAs that do support OCSP may need to log every GET/POST request for OCSP responses, and keep this data for at least 2 years.
The requirement for CAs to monitor OCSP requests is the product of a different time, when thinking around OCSP was very different. As privacy concerns and other structural weaknesses move the community away from its position on OCSP, it no longer makes sense to include requirements for CAs to watch and record OCSP requests.
Ballot SC-063 v4 made it optional for CAs to provide OCSP at all. (We recognize that there is still a root program requirement that pragmatically prevents CAs from eliminating OCSP, but within the scope of CABF requirements this is a critical change.) For the BRs to strongly recommend (via this SHOULD requirement) monitoring OCSP is incongruous and out of keeping with current thinking.
Even if we did want such monitoring to take place, any such requirement would present serious and perhaps insurmountable technical challenges:
For a typical OCSP responder that is only aware of unexpired certificates, it's impossible to tell the difference between an "unused" serial number and the serial number of an expired certificate. To disambiguate would require the ongoing cross-referencing of OCSP responder logs against the CA's cert issuance logs, requiring additional code development and maintenance and significant production overhead.
Furthermore, as most OCSP services are fronted by CDNs, there's no guarantee that the CA will even have access to the full OCSP request logs. If the CA can't enumerate all the IP addresses of OCSP clients that send requests for "unused" serial numbers, then this vastly diminishes whatever value we attribute to this monitoring requirement.
Our proposed changes are available for review on https://github.com/cabforum/servercert/compare/main...XolphinMartijn:servercert:LoggingRequirements <_blank>.
With this email I’m hoping to receive feedback and thoughts on this proposal.
Servercert-wg mailing list
Servercert-wg at cabforum.org <_blank>
Daniel Jeffery | TLS
fastly.com <_blank> | @fastly <_blank> | LinkedIn <_blank>
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 8254 bytes
Desc: not available
More information about the Servercert-wg