Slashdot Mirror
Symantec Subsidiary Thawte Issues Rogue Google Certificates
New submitter jack_babylon writes: On September 14th, Symantec's subsidiary certificate authority Thawte accidentally released a "small number" of " "inappropriately issued" security certificates, apparently intended for internal testing only. However, the fact that these were logged in the wild by Google (and, apparently, DigiCert) seems to indicate that they escaped the lab, at least far enough for a false google.com cert to raise the appropriate red flags. This sounds similar to the recent acts of poor judgement that got CNNIC's certs removed entirely from Firefox and Chrome, if more limited in scope and more quickly addressed (through, among other things, termination of some Symantec employees). (And like all reports one hopes go away quietly, these were released in the dead of a Friday night — h/t BoingBoing for noting this news.)
That's some very suspicious "testing", kids.
What are you up to?
Perhaps in bed with some three-lettered thugs?
Not the GP poster, but here goes:
The ideal situation is that the Certificate owner generates a signing request and has that signed, so the original key does not go outside the certificate owner.
However, there is nothing in the current setup to prevent a certificate authority from generating a request in the name of any domain and signing it. That's what appears to have happened here.
The real question is 'why?'. The explanation ("testing") doesn't pass muster. Someone would have to deploy these certificates on a service that was either a Google property or was masquerading for a Google property. Does Google outsource the deployment of certificates? I would doubt this very much, which suggests that this wasn't so much an accident as the influence of a TLA.
The real "Libtards" are the Libertarians!
Any trusted certificate authority can issue certificates for ANY domain. This is the trust aspect that is required in a PKI.
Your browser gets a list of trusted root certificates and will accept any valid certificate issued by these CAs. On my windows 8 box there are 53. Any of these providers could issue certificates for any number of domains.
The failure here is that Thawte allowed those certificates to be issued for ANY reason.
Google is their own certificate authority and likely has no need for a relationship with Thawte.
From the summary: "...termination of some Symantec employees..."
Is this the first time that individuals were held responsible for online negligence? What happens to a CEO or CIO when data on millions of people slips out due to negligence? Has anyone ever been fired before (not just a flunky, but a responsible executive)?
The penalties for corporate irresponsibility are so small that there is no incentive to do the right thing. Actually, this case may be an exception because both Thawte and Symantec have a reputation to protect- they might actually fire an executive. The question remains (and you can ask the same of Wall Street criminals)- when has any executive ever paid for this kind of negligence?
...omphaloskepsis often...
Security company dicking up the one thing they are supposed to be good at.
My brother had to cancel his honeymoon last month. He had his vacation time denied by Microsoft. The thing that has made him so angry is that since then several Indian coworkers have been allowed two week or longer vacations.
Sure it's not a perfect fix but publishing the signatures of your ssl certs in DNS would care care of a lot of this low hanging fruit. A standard for cosigning your certs and pinning that cert would also help.
The end effect is needing to break multiple vectors not any of a multitude of root level CA's.
No sir I dont like it.
It's not that Google or Thawte have failed to correctly revoke certificates: it's that far too many people, at far too many sites and with far too many technologies, do not actually keep their signature authorities up-to-date. Because these people don't update signature authorities, they are unable to verify numerous valid certificates. These people then simply set their automated procedures, or make it their personal practice, to accept invalid certificates.
The notable case of this I saw recently is RHEL 5, where the signature authority information in the /etc/pki files managed inside the openssl packages, and updating openssl on a live server is likely to cause fascinating problems for an old, stable, production server. RHEL 6 sensibly put the root SSL keys in a separate package, but it's certainly not an unusual problem.
It's all good now. Employees have been fired and some guy on a Symantec blog said the internet was never at risk. We all can relax and enjoy life.
lucm, indeed.
Your browser gets a list of trusted root certificates and will accept any valid certificate issued by these CAs. On my windows 8 box there are 53. Any of these providers could issue certificates for any number of domains.
Worse those providers can issue "intermediate certificates" which also have the power to issue certificates for any number of domains. They can and do issue those intermediate certificates to third parties. So the list of root certs in your browser is not a complete list of entities who can issue certs your browser will trust.
There was recently an extension added to allow intermediate certs to be limited to certain ranges of names but that only helps in clients new enough to recognise the extension.
There was also recently an extension added for "key pinning" which makes bogus certs less useful.
Google is their own certificate authority.
At least when I go to google and check the cert I get a cert that has a google intermediate and a geotrust root. I don't see any evidence of name constraints on said intermediate cert though :(
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
The problem is that any of the many entities your browser trusts can create a valid certificate for any domain and the browser will just accept it.
What we need is to move away from CAs and adopt a new system for storing the information needed to make a web connection secure. Storing keys in DNS and using DNSSEC to secure that is one option. And there are others (although I can't actually remember any of them off the top of my head).
If you have a situation where its impossible for anyone other than the actual owner of the domain to store a key, its not possible for a rogue CA (or a hacked CA ala DigiNotar or one that has been co-oped by a government or intelligence agency) to issue a bogus certificate or a bogus public key.
What I dont get is why there is no real interest from the people who came up with these alternatives to push them particularly hard and why there is basically zero interest from the people and entities who write the software that the web runs on (browsers, servers etc) to make any moves towards using these new systems.
I live in Iran, most probably they were issued for my government, this is the most practical solution here.
Moderating "-1, Disagree" is simple censorship. Have the guts to post your opinion. -- Spazmania (174582)
Two other strategies are certificate pinning and certificate transparency. For pinning, you declare that only a certain intermediary CA (or root CA) may sign certs for your domain. So Google basically declares that all *.google.com certs must be signed by the specified Google CA. This information can either be hardcoded in the browser (for major sites) or relayed the first time the browser contacts the domain. So with hardcoded pinning, only Google can sign their certs. With http-pinning, another CA could only spoof the cert if the user had never loaded Google before.
The other initiative is certificate transparency. Basically it's a public log of all certs issued by participating CAs. (Non-participating CAs shouldn't be trusted by default). Transparency would have prevented the false Google cert that happened a year or two ago because the unrestricted signing cert granted to the customer would have been immediately flagged as suspicious. If an invalid cert is signed directly by a trusted root CA, it can be detected immediately because Google would be monitoring the public logs and see a cert being issued for their domain. Smaller companies which don't want to directly monitor logs themselves can sign up for a notification service. Suppose I offer such a service and you are my customer. Any time a cert is issued for one of your domains, I'd call you and let you know. You could then take immediate action if it was improper. A company offering such a service would probably also offer assistance in handling the situation.
As for browsers, I should be able to remove Thawte from the trusted chain
Go ahead. In Firefox, hamburger->options->advanced->certificates->view certificates. Find the two headings for Thawte and set all of their entries to "distrust". I've no idea exactly how much of the web will stop working correctly after that, but it's not hard to do.
I should be able to configure a warning if a domain has changed its certificate authority chain since the last time we saw it.
You should, and I'm sure there's some kind of add-on or setting for that, but I don't know what it would be off the top of my head.
It is pitch black. You are likely to be eaten by a grue.
Certificate Patrol would be that plugin : https://addons.mozilla.org/addon/certificate-patrol/
But I cannot understand why it is used so rarely. There also used to be DANE Patrol which used to do the same thing while being able to handle multiple certs for 1 domain (think Google) made by Czech NIC, but it is not developed anymore and they recommend not to use it.
Security requires both encryption and trusted identification. The first is easy, the second is what the CAs are for.
Certificates are there for security,
Yup, and you can tell how well they work for that by seeing how good a job they're doing in stopping phishing, malware, and spam.
Certificates are there to make money for commercial CAs because web sites are forced to pay them protection money to turn off the browser warning messages. That's all they do.
What needs to happen until a corporation is terminated?
That is the main issue here. As human beings, we understand there are limits to what we can do before we face really serious consequences. I mean jail time, not monetary punishment. Money is simply an expense. It might hurt, even hurt a lot, but it is not on the same level as being locked up.
Where is the jail time equivalent for corporations, and why do we continue to believe that we can somehow control them without it? To take back control of our worlds from corporations running amok over it, we need this.
To fire a number of employees means something very seriously went wrong. It also means the corporation allowed it to go wrong. This could be rogue employees the way someone robbing a bank with your car had lied to you when borrowing it, saying he needs to visit his ill grandmother urgently. Or it could be that you gave a gun to an obviously unstable kid when he said he's going to school and he's angry. You really should have at least asked a few more questions before handing over the firearm.
So what will happen to Thawte in response?
Assorted stuff I do sometimes: Lemuria.org