The fine folks over as SANS ISC tweeted this out last week:
The GlobalSign CRL just "exploded" with > 54k revocations today. removed it from our graph for now. https://t.co/JZdDu7ulK7— SANS ISC (@sans_isc) April 16, 2014
Now, Netcraft monitors more CRL lists than the ISC does but their revocations-per-hour chart shows a similar pattern:
The second Netcraft article does enough of a good job stating most of the problems related to transport security & integrity that we won’t go into the issues here. It’s a bit sad, though that:
- Safari & Firefox do “best effort” OSCP by default
- Google has it’s own way of dealing with browser-based certificate checks
- The use of CRLs by issuers is inconsistent at best
- OCSP stapling (which deals with some of the privacy concerns of OSCP) is not ubiquitous
- Soft-fail is still ‘a thing’
reasonfield for each revocation is rarely filled out, making the whole thing (in
@hrbrmstr‘s opinion) less than useful
If you want to play along at home with Netcraft & SANS ISC, you could go the route of hand-grabbing each of the 79 certificates on the SANS list, running:
openssl crl -in <filename> -inform DER -text
individually on each one and processing the output, but it might be a bit better to use our CSV version [GH] and write a script to grab the files (which is simple enough to not put here).
Rather than rely on the
openssl command-line tool, you can use the
PyOpenSSL library to process the files, which will make it easier to grab the reason for revocation:
import OpenSSL import sys # pass in the PEM CRL file CRLFile = sys.argv # read it in with open(CRLFile, 'r') as crlfile: CRL = "".join(crlfile.readlines()) # create the CRL object parsedCRL = OpenSSL.crypto.load_crl(OpenSSL.crypto.FILETYPE_PEM, CRL) # get all the 'revoked' entried revoked = parsedCRL.get_revoked() # print out each entry for entry in revoked: print f, entry.get_rev_date(), entry.get_reason()
You’ll need to do a translation step first from DER to PEM:
openssl crl -inform DER -outform PEM -in revoked.crl -out revoked.pem
and, will probably want to validate the format ahead of time since some files are actually PEM files to begin with.
We’ve teased out some graphs on Twitter the other day and could have just posted end-to-end code here. Rather than do that, we’re issuing a DDS reader challenge! Winning is simple! Just provide a complete solution to the following requirements and you’ll be entered to win a signed copy of our book. All you have to do to enter is send an e-mail to
contest at datadrivensecurity dot info with a link to a GitHub Gist (any language/platform within reason), an iPython Notebook (via the iPython Notebook Viewer) or RPubs that does the following:
- has a script/program that uses the CSV provided (or a better one you can find or create that has even more issuers) that downloads the CRLs and generates a data file or database records that have (at a minmum)
reasonfields (you can include more if you want)
- has one or more additional scripts/programs that performs analyses and generates charts with explanations that tells the story of what CRL revocations looked like before and after the Heartbleed “crisis”. At a minimum, you must show by-day, by-hour, by-issuer and by-reason views
- extra consideration will be given to submissions that are repeatable (i.e. not one-off scripts)
Once we receive the e-mail, you’ll get a confirmation response (ping us on Twitter if you don’t get a response as it’s a GoDaddy forwarder).
Make sure you don’t blather out the URL you provide to us before the contest! We’ll publish all working entries in the post that announces the winner. If you already have a copy of our book or acquire a copy to help with the contest, we’ll work out another book-prize (preferably from Wiley Press :-) with you.
This contest will run until Sunday, April 26, 2014 23:59PDT and the winner will be announced by Wednesday of that week.
Questions & clarifications can be asked of either Bob or Jay on Twitter or in the comments.Tweet