Improving revocation : will Let's Encrypt support OCSP Must-staple?

Feature Requests
41

The CPS has been updated, yes.

42

CP: ."id-kp-emailProtection [RFC5280] maybe present. Other values shouldnotbe present"
Is this an hint that there maybe come s/mime certificates too ? :slight_smile:

And an typo on the head of page 62:
“Curve P-521” “ansip384r1 ::= { iso(1) identified-organization(3) certicom(132) curve(0) 35 }”

43

Don't think you can say anything about that. That text isn't new and was already incuded in the first CP of 2015.

44

Now suche.org works but I had to manually delete old HPKP to see it working…
You should double check your HPKP configuration.

45

depends on whether you are even able to do that in the first place, epending on your browser.

remember that (sadly) HSTS and HPKP is nothing the user can circumvent

46

And how can I create a CSR with the Must-Staple extension?
With OpenSSL?

(Is a special OpenSSL version required for this? If so, what is the minimum version where it works?)

47

See this post on how to create a CSR (also see here for a full CSR example).

“1.3.6.1.5.5.7.1.24=DER:30:03:02:01:05” is the part you need to add to the “req_extensions” section (you need to type “1.3.6.1.5.5.7.1.24” for older openssl versions since the option “tlsfeature=status_request” is not mapped yet automatically).

But beware when you use this for other services despite web servers, e.g. postfix and dovecot do not support ocp stapling, but e.g. Thunderbird checks whether a stapled response is included.

2 Likes
48

If you use the official client: since 0.7.0 there’s a new switch: --must-staple

49

BTW: Do you know any testing tool, which shows whether OCSP Must-Staple is used/working?

50

I don’t know any tools, but you can check if the certificate has the must staple attribute. And just after installing the certificate, the first request will probably not be staple, so should failed with Firefox.

1 Like
51

Interesting. This is exactly what it did.
Additionally you can also see it in the certificate details view of Firefox under Certificate -> Extensions. There is an "object identifier" 1 3 6 1 5 5 7 1 24 with the content 30 03 02 01 05. That's what you've set in the OpenSSL config file.
BTW: You can also see this in the certificate transparency log.

1 Like
52

This behaviour (first request fails due to lack of OCSP staple) is a defect in the web server. Make sure to complain to the people who wrote it and don’t take “it’s supposed to do that” for an answer. Imagine if they said “It’s supposed to do that” for giving a 404 on every tenth URL access. Lazy is what it is.

What web servers with OCSP-stapling turned on should do is spin up, notice they have one or more SSL certificates and check a permanent (e.g. disk) cache to find a response for each cert. Do they have a new enough OCSP response cached? If so, start serving that immediately. Otherwise, spin up a sub-routine to fetch a new OCSP response.

Apache’s current attempt is pretty feeble. They have no permanent cache, if you stop Apache to patch something, when you restart it their OCSP “cache” is empty and you’ll get empty results for a while. That’s the sort of thing you’d expect in some “proof of concept” demo, not a real web server. Responsible people aren’t going to be able to turn on OCSP-Must-Staple for an Apache server until they do a much better job of this and that starts with feedback.

1 Like
53

In my case I could reproduce it with nginx 1.10.0.

But I found a guide with some workarounds:
https://unmitigatedrisk.com/?p=241

1 Like
54

BTW SSLLabs works on implementing a test for this feature.

2 Likes
55

Just for information, in Firefox, an certificate with Must-staple seams to breaks HPKP ( see Bug 1278041 )

1 Like
56

BTW: You can vote for (fixing) this bug on Bugzilla. :smiley:

2 Likes
57

On my page suche.org i do not get an hpkp error and use the must staple.

58

@tlussnig I do get the same error in the console with firefox 46.0.1 about your HPKP.

59

Hi, now i also see it in the console log.

  1. It is interesting that an HPKP error does not cause an visible security alert.
  2. SSL-Errors should normally be reported. But this is also not triggered.
  3. SSL-Labs see no error on the page.
60
  1. When an HPKP header is ignored, it doesn’t cause a visible security alert, and the consequence of the bug is that HPKP is ignored.
  2. As the header is ignored, no report…
  3. The header is correct, so no error to show !
1 Like