I’ve been looking into MFA solutions for our Windows domain the last couple of months. I read Suzan Bradley is about to publish an article on the subject and I can’t wait to see what she has to say.
So the last couple of months I’ve been testing DUO, ADselfServicePlus and Userlock. With Userlock, I encountered something that in my opinion, could cause a security risk. All three solutions use an Agent to be installed on a PC. You log on using name and password and then the Agent pops up requesting a code, generated by Google Authenticator or alike. With Userlock, logging on is a bit different compared to the other two. After entering username and password, Windows starts downloading the user profile and when finished, the Userlock agent pops up requesting a security code. And that poses a security problem, for which I made up a quite convoluted and hypothetical scenario:
Suppose a company where most users are added to the group Domain Admins. They use MFA, so there’s no immediate harm if a hacker has got hold of someones login name and password. That hacker would need the token or authenticator as well to get access to the domain. One day, a hacker indeed has got hold of someones login name and password. He manages to hook up his pc to their network and to his surprise, he’s able to join his pc to the domain. That’s a nice bonus, he thought, I’m already Admin so this is a simple job. Then he tries to logon, but darn, they use MFA. But he does note the user profile is downloaded to his machine. So lets have a look what this user has stored in his profile. My documents, My pictures, the whole lot is there. And there’s another bonus – on the desktop there’s a text file with the Userlock emergency key, placed there when enrolled with Userlock. The user didn’t see a need to store that file in a safe place since they are protected with MFA.
That’s the scenario I came up with. I contacted Userlock about this and they fail to see the problem. To enlighten them, I did some more testing. Then it dawned on me the scenario above is way too complicated – it’s far more easy to get access to that companies network. Thing is, the MFA Agent ONLY kicks in when installed on a machine. Machines without the Agent simply don’t do MFA at all. So in the above scenario, the hacker doesn’t need to join his pc to the companies network; the only thing he has to do is find some way to connect to their network, open Explorer, type the name of a server / share, Windows pops up a window to enter user credentials, and you’re in. No need for Admin-permissions as well. So I talked about this new insight with Userlock and their answer is their solution is to protect machines.
That’s not the way MFA should work, in my opinion. I want MFA to protect accounts. So no matter what’s the source used to access our domain, as soon as you have to enter your user credentials, there should always be the second authentication method. Not able to supply a code or token or whatever you use as second method? Then access is denied.
Now I wonder – do all MFA solutions work like this? If so, then they are completely useless??