InfoSec Insider

Kerberos Authentication Spoofing: Don’t Bypass the Spec

Yaron Kassner, CTO at Silverfort, discusses authentication-bypass bugs in Cisco ASA, F5 Big-IP, IBM QRadar and Palo Alto Networks PAN-OS.

Authentication is the front gate to security systems, so if you bypass it, you can pretty much do whatever you want. You can log in as an admin and change configurations,  access protected resources and gain control of appliances to steal sensitive data from them. For these reasons, the authentication protocols used by security systems must be flawless. But in security, there’s no such thing as a flawless system, and implementation errors can lead to hazardous security vulnerabilities.

And that’s exactly what we discovered when analyzing four different security systems –  Cisco ASA, F5 Big-IP, IBM QRadar and Palo Alto Networks PAN-OS. All were vulnerable to bypass exploits because of the way they implemented the Kerberos and LDAP authentication protocols.

Infosec Insiders Newsletter

To understand these vulnerabilities, one must first understand how the Kerberos authentication protocol works. It consists of three exchanges:

  • In the Authentication Service exchange, a user logs in to a client via a username and password to authenticate to an authentication service, which resides in a Key Distribution Center (KDC). In return, the authentication service provides a Ticket Granting Ticket (TGT), which is used in the next exchange.
  • Then, in the Ticket Granting Service exchange, the client presents the TGT to a Ticket Granting Service (TGS), which also resides in the KDC, and requests a ticket to access a particular service. The TGS returns a service ticket.
  • Finally, in the Client/Server exchange, the client presents the service ticket to the server, which verifies it, and allows the client access to the service.

The Kerberos protocol is solid. It was developed at MIT and provides Single Sign On (SSO) for many large companies.

Poor Configuration Can Lead to Attack

The four security systems we mentioned earlier can be configured to use Kerberos without its SSO capabilities. Instead, the user is prompted for a username and password when logging in, and the system asks for the TGT. In other words, the security system serves as both the client and the server.

One might think – if the client and server are the same, why do I need the client/server exchange? The password is verified during the Authentication Service exchange, so that should be enough. This thought process sounds legit, only they forgot the first rule of fight club: Don’t deviate from the spec.

KDC Spoofing Attacks

In fact, this is not legit at all. Some would say it’s even illegitimate. Neglecting the Client/Server exchange can result in a KDC spoofing vulnerability. Let’s look at it from an attacker’s perspective:

I want to authenticate to the system, but I don’t know the password. So maybe I’ll try to authenticate with a password of my choice – it’s most likely the wrong password, but the idea is to trick the system into thinking it’s the right password.

The system, being the Kerberos client, will reach out to the KDC to request a TGT. If this message is actually processed and answered by the real KDC, the attack will not work, because the KDC will notice that the password is wrong, and simply deny the authentication. But if we hijack the communication between the system and the KDC, then we can do some really evil stuff.

What if I pretend to be the KDC, and instead of returning an error, I return a fake TGT?

Well, in that case, during the TGS exchange, the KDC will reject my TGT. But I can hijack the TGS exchange communication as well and return a fake service ticket. Oh, but then the Client/Server exchange won’t work, because the server will reject my fake service ticket.

Then again, these four security vendors didn’t implement the Client/Server exchange at all. So I can just log in with my fake password to all these systems.

In reality, discovering these vulnerabilities was a bit more complicated than that, because each of the products implemented the Kerberos protocol a little bit differently, and this required changes to the original attack pattern. I encourage you to read more about each vulnerability in these blog posts – CVE-2019-4545, CVE-2020-2002, CVE-2020-3125, CVE-2021-23008

It’s interesting to see how these implementation errors can be overlooked by so many developers. We responsibly disclosed them to the four vendors and worked with them to issue fixes, so the users of the systems are safe now, but there is a lesson to be learned here.

Don’t deviate from the spec!

Acknowledgements:

The said vulnerabilities were discovered by Silverfort researchers Yoav Iellin, Dor Segal, Rotem Zach and me. The Big IP vulnerability was a joint effort with Thierry Van Steirteghem, who worked at Exclusive Networks at the time of discovery.

Yaron Kassner is co-founder and CTO at Silverfort.

Enjoy additional insights from Threatpost’s InfoSec Insider community by visiting our microsite.

Suggested articles

Epik Confirms Hack, Gigabytes of Data on Offer

“Time to find out who in your family secretly ran … [a] QAnon hellhole,” said attackers who affiliated themselves with the hacktivist collective Anonymous, noting that Epik had laughable security.

Discussion

Leave A Comment

 

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Subscribe to our newsletter, Threatpost Today!

Get the latest breaking news delivered daily to your inbox.