Imagine a group of researchers planning to speak at a conference regarding a previously undiscovered vulnerability present in most homes that would allow a thief to rob your home of its valuables with complete ease. You would probably be interested in hearing what they had to say so you could take the necessary precautions to protect your home.
Now imagine when they presented their findings, they went on to state that it was incredibly easy to do, so long as you left your front door open and also provided them with the security code for any alarm systems. You would probably find this implausible and simply the proliferation of fear, uncertainty, and doubt.
That’s precisely what happened last week at the well-respected Black Hat security conference in Las Vegas when researchers from the Israel Institute of Technology and Advanced Defense Systems, Ltd. presented their findings of a serious vulnerability present in OSPF. So serious in fact, the researchers stated the only way to properly mitigate the threat, short of fixing the protocol, is to switch to another routing protocol such as RIP or IS-IS.
The researchers went on to present their research of two previously undiscovered methods of poisoning routing information within an OSPF domain such that bogus information could be propagated and used for a multitude of malicious purposes.
The first attack allows for the creation of a remote false adjacency, whereby an unsuspecting victim router goes on to form an adjacency with a phantom router. Once this adjacency is formed, the attacker can then flood the victim router with bogus Link-State Advertisements (LSAs).
The second attack allows the flooding of what are known as Disguised LSAs, whereby a victim router would accept spoofed LSAs which it believed were generated by another router within the OSPF domain.
While both of these previously unknown attack vectors might raise some eyebrows and create cause for concern, the researchers made some pretty big assumptions regarding the ability to perform such an attack. The assumptions were that in order to perform such an attack, one would first need to insert a router into the network, and once this was accomplished they would then somehow need to obtain the MD5 authentication key that is widely used to secure the communications channel between OSPF neighbors.
Without a doubt, there is some merit to what these researchers have found and to a protocol junkie like myself I am always interested to learn the inner-workings of these protocols. However, what they’ve completely overlooked is the first tenet of information security which is to ensure the physical environment is properly secured. In most environments, gaining physical access to the network might prove difficult enough. Nonetheless, if an attacker were able to insert such a device, the net effect would be that such attacks would be short-lived as it would be fairly trivial for a network administrator to isolate the offending device and remove it from the network.
Furthermore, and perhaps the biggest barrier to entry, they’ve failed to recognize that the likelihood of someone obtaining the authentication key. While MD5 is known to have some serious deficiencies and vulnerable to collision attacks, it still represents a pretty high bar for the would-be attacker. And various vendors are already rolling out HMAC-SHA for the authentication function within OSPFv2.
In addition, the researchers admitted that in their testing of the first exploit, the victim router tore down the adjacency to the phantom router after 125 seconds because the victim router did not receive the proper acknowledgement packets from the phantom router. Once again, the net effect is that if the attackers were able to meet the first two assumptions, their attack would still be relatively short-lived.
While OSPF is widely used throughout the Internet as a whole, you probably have better odds of winning the lottery than actually experiencing an attack such as these researchers have demonstrated. The researchers presented some pretty bold claims, but at the end of the day didn’t really present anything that wasn’t already widely known about the vulnerabilities inherent within the OSPF protocol given the assumptions the researchers made. Their assumptions are based on two barriers to entry, which doesn’t make it impossible, but nonetheless incredibly difficult and highly implausible. In my experience, attackers tend to move onto far easier targets.
There are a lot of things to worry about when it comes to protecting your corporate assets, but in my humble opinion, this is not one of them. So rather than ripping out OSPF as these researchers would have you believe, you might perhaps be better served by taking a second look at your environment and ensuring the physical premises are properly secured.