Gregg O’Brien is a Microsoft Premier Field Engineer from Canada. In this post he talks about the ‘certificate explosion’ phenomenon and suggests a way to mitigate it.
We live in some very exciting times – we have so many devices to choose from: desktops, laptops, tablets, hybrids/convertibles, ultrabooks, netbooks and smartphones. Each of these devices offer their own unique benefits and features, so much so that it’s not uncommon to find people carrying 2 or 3 devices now!
As with all super cool technology though, IT pros will always find some challenge waiting at the bottom of that pile of coolness. In the case of multiple devices in an enterprise, a common problem is enrollment of certificates. Not so much a problem of acquiring certificates, but the problem of users acquiring too many certificates.
A Microsoft Active Directory Certificate Services infrastructure on Windows Server 2008/2012 is implemented with auto-enrollment capabilities. Users with accounts in Active Directory login to the domain and the auto-enrollment policy enrolls the user for a certificate tied to their account. The certificate is downloaded from the certificate authority and is stored in the user’s certificate store on the local computer. So far so good….
Now for the ‘wrench in the gears’: the same user logs into another computer with the same user account and because the certificate store tied to that user account is empty on the second computer, the user receives a new certificate with a different private key. This behavior repeats on every computer the user logs on to. At first this might not seem like such a big deal. But this actually presents a few potential issues:
So what can we do about this? Well, the good news is, unlike most complex problems in life this one can be fixed by checking a few check boxes:
Now when users enroll for a certificate, the client will first check it see if there is a certificate in Active Directory. If there is, rather than issuing a new certificate it will reuse the certificate that has already been issued. Another technical challenge vanquished!
Posted by Arvind Shyamsundar, MSPFE Editor.
Thanks for this article!
Stupidly we overlooked the setting to avoid multiple certificates and now we have the problem that several users have more than 1 certificate that is published in AD for email encryption.
What is the best way to cleanup? Check what certificate will be used as default one and delete the other?
Is it surely that the default certificate will be used everytime a user send an encrypted mail?
What is the best practise?
Thank you very much!
@toasti - best to get someone in or open a case with Microsoft to evaluate it, there are lots of different components to possible solutions. They involve complexities like credential roaming, roaming profiles, certificate distribution, key storage location, and more. You may have different "default" certificates on different devices.
Thanks Alan for your help.
I see, not so easy but I didn't thought it will.
The checkbox as described above is set since yesterday, is it a possibility just to wait until they expire and then there's only one valid certificate in the store due to the setting to do not enroll multiple certs?
But as I understand, I still must care of that the user has all his old private keys on the machine to read oly emails, right?
At the moment we do not use roaming profiles and credential roaming, but credential roaming sounds like the one we need.
Hi, I followed the exact same steps as above but still when a user is logging into multiple computers, it is generating individual certificate for each computer. Please help
Hi all, Our CA server was configured to avoid user enroll for multiple certificate and to archive users' encrypted private key in server ("Archive subject's encryption private key" in Request Handling tab). Users have to set password to protect their private
key stored in local computer. Our problem is how to protect Users' certificate( signatures) however they lost their AD account. This mean, when Attacker know AD account of an user, attacker could log on user's computer, delete user's certificate in local computer
(password protected) then request a new certificate. we want to configure CA server to required attacker provide password which user set at the first time request certificate for each time re-request certificate. Please help!
Assuming this also requires credential roaming? As in out environment this does not work.
Yes and no. For this to work with users using lots of computers, you'll need to ensure the private key for the "one" certificate are stored on each computer the users use. Which means, in general:
-credential roaming, or
are needed to keep that private key material consistent across hosts.
"For this to work with users using lots of computers, you'll need to ensure the private key for the "one" certificate are stored on each computer the users use. Which means, in general:
-credential roaming, or
are needed to keep that private key material consistent across hosts."
This article is badly incomplete without telling this tiny little bit missing part of the story to the readers!
Attention to everybody out there reading this article!
Will this setting prevent users from re-enrolling after their certificate has expired?
Won't an expired certificate remain in Active Directory and thus prevent the user from being able to renew?
Any thought about reuse if the certificate requester is a service account but we are associating the cert with a user (aka, MDM like AirWatch)? Trying to use the same AD user cert for both a PC and a mobile device.
Once certificate is created on first user's login, and attached to its AD account, this setting (do not automatically reenroll a certificate if a duplicate exists in AD) prevent a new certificate to be created on another workstation when user log in. Is there
any way to have the first certificate (attached in AD user's account) copied on the second workstation instead ?
Thanks for your replies !
Hi François ,
I think you will find that the auto enrolled Public Certificate itself IS published in AD meaning that if you open up "certmgr.msc", you check out the "Active Directory User Objects", you will see all the certificates issued to the user in there. However, there
are no references to the private key - I believe this is stored on the Local computer and if you check out the "Personal" folder, you should see one certificate in there that matches one of the AD certs, but will have a key on the icon. Opening up the Certificate,
and you should see a note at the bottom saying "You have a private key that corresponds to this certificate". Without the private key, you cannot decrypt anything that has been encrypted by a third party with your public key. The point of the Private Key is
that its supposed to be private, so publishing this in AD is probably a bad idea.
Our issue is that we us a LOT of RDP sessions to AD joined servers, and users are also issued with certificates for every machine they login into via RDP. I can see on real way of figuring out what machine was used to generate what certificate. I really want
to get away from all these multiple certs, but of course would run into the same problem.
I would love to hear what the original writer as a"Microsoft Premier Field Engineer" would say on the subject.
And to follow up, as this was posted over 2 years ago, I can't see anyone actually following up.
Thanks for your feedback.
Correct: i can see certificates attached to my AD user in "AD User Object" store. But i can't use them on web sites (for authentication, i don't need to use them for decryption).
By the way, i opened a case to MS support. First answer was to enable Creadential Roaming using GPO (User Configuration / Policies / Windows Settings / Security Settings / Public Key Policies / Certificate Services Client - Credential Roaming Settings). Tests
are on going...