RSG Model Works

Check out https://plus.google.com. It’s got a certificate for “*.google.com”. Wildcard certs may be the “store the used control rods in the attic and forget about them” technical trick of the certificate world. But wait, it gets better. This was issued by the “Google Internet Authority”. This presumptuous name describes a Certificate Authority, operated by Google (a/k/a google.com), that is in turn signed by a GeoTrust root.

Uh, guys, the point of certificates was to introduce a “trusted third party” (see Wikipedia definition or use BING to search for the term yourself…)

When the company running the web site issues the cert for a public TLS-protected website, the point was to be able to trust it because SOMEONE ELSE was the trusted third party. That’s why they are the THIRD party (not the first party, the browser user, called the “relying party” in certificateSpeak, or the second party, the private key holder of the site being accessed via “https”.)

Other sites do this. Akamai has in the past (and may still) practice this “I’m the second party and the third party” stunt.

Is this bad? Yes, this is bad. There’s no trust here. You’re only trusting the web site operator. If they are compromised, or go rogue, you’ve got no recourse. Revoking the certificate is no longer a defense. Trusting the root is no longer a defense. And, it implies these retail certificate authorities will take money for all sorts of crazy non-trust-delivering practices.

Check out this . Note the hostname mismatch (it’s got a GeoTrust cert for www.adgrafics.com. Note the WEB TRUST seal in the upper right corner. Click on that, let Chrome kindly translate (appologies, I don’t read Russian or Ukranian.) Note the seal is from “https://webtrustukraineseal.com” (confused yet?) and THAT says “Verisign Trusted” (Verisign != GeoTrust.) Note also the default https://www.certificatesigningrequest.com:8443/ Plesk self-signed certificate.

So… you should trust these certificates? From several different CA’s, and from a site that uses self-signed certificates? Not to pick on Symantec’s Ukrainian trading partners… but really, are we supposed to trust these certificates? When the vendors are this sloppy?

After your enterprise has grown up enough to really need electricity all the time, uninterruptable power becomes a necessity.

In the 70’s if you drove through Waltham Massachusetts on Route 128 you could point out the large enterprises.  They all had a Cat diesel engine mounted on cinder blocks behind the building, set up to run a generator.  You’d put it six feet up on cinder blocks, behind the building so that for example a vehicle in the parking lot wouldn’t have a chance to crash into it.  I noticed these because to me a Caterpillar product was a farm tractor sold as far back as the 1930’s (my father sold three rail freight cars full of Cat D-series tractors one season.)  They looked quite silly to me, until I realized the IBM 370’s we were using would crash horribly if those Cats weren’t out in the back yard.

My point is that’s an OLD STORY.  Check out that “one page tsunami plan” news item on the internet.  Look for the AP version here. Check out picture 4 of the TEPCO-released images in the side bar (also below). That’s water gushing into the basement where the dieslel generators for the coastal nuclear power plant were deployed.

Yeah. After the boss buys you a UPS (because you did pass D/R 201), put it somewhere sensible.

Years ago, at the dawn of the dot-Com age, when crypto was cool and Distinguished Names were already an arcane concept, there was a story, let’s be kind and say it’s an urban legend, about root keys.  In the early days you bought a BBN Safekeeper.  It kept the RSA private key safe.  It had a battery backup on the memory it used to store the keys (remember this would have been 1980’s tech.)

There was this story about how American Express bought a Safekeeper but forgot to change the batteries.  I’m not sure it’s true but it does point out the need for the key operator to follow policy and use the “split the key and save the parts in separate places” features of modern HSM solutions.

More generally, you should buy a UPS.  Or at least make sure someone’s making sure your expectations about continuous clean in-budget power are met.  Buy a UPS, make sure you plug into the “special” power strip in the Colo, confirm the D/R plan is NOT on your task list, or somehow think about it.  At least think about it for a moment.

This is DEFINITELY not the PowerPoint slide you want shared at your next project postmortem meeting where you discuss poor estimates of project risk.

On the other hand, the next postmortem I have to attend is one of those “…and THAT is why I was standing in the data center in front of the Cisco switch, grumbling about change management” meetings and I SO am putting this image in my PowerPoint deck as a warning.  Something about letting those cowboys out in the Gulf play fast and loose with the tech…