Roger's Information Security Blog

Here are some notes from Shmoocon day 2. Today was a return to the traditional Build It, Break It, and Bring it on tracks. Here are some notes/summaries from the sessions I attended. It was another fun day.
Active 802.11 Fingerprinting, Bratus, Cornelius and Peebles
How can you identify if an access point is legitimate or rogue? Does two way RSA crypto solve the problem of a rogue AP? The speakers would argue that if you are communicating with a rogue AP, the use of certificates could actually cause more information to be given away to the rogue. You could certainly be exploited in your communication as well if your wireless drivers have vulnerabilities.
Just as with OS fingerprinting through TCP, the wireless protocol can be abused to send unexpected traffic to the AP and fingerprint how it responds. They built a tool called Baffle using Ruby to perform this test. They were able to verify that the access point was using the driver that is expected.
If you’re expecting a linksys AP and I set up a rogue linksys AP, this isn’t going to help you, at least from my understanding of the talk. An audience member asked if this could be used with adhoc (client-to-client) connections as well. It cannot be used for that because the APs are much more chatty and have more negotiation.
The remainder of the time was a presentation on access point hiding. I did not catch the presenters name. Basically anything that has some room inside and has sufficient power could be refashioned to contain an AP. This assumes that you need to be stealthy about placing a rogue AP in the first place. The take home for me from this section of the talk was the question, “if an AP enabled itself at 2 am (either to let the hacker in, or to move some data out) would you catch that.”
Smarter Password Cracking; Weir, Glodek
Not a lot new here.
Password cracking is getting tougher. Sometimes users are forced to pick better passwords. Often developers are throwing in a salt or hashing multiple times. A salt makes a precalculated table attack difficult. Multiple hashes attempt to increase the calculation penalty when trying a offline password attack. For example while Word’s password mechanism was once trivial to break, Word now uses 5000 SHA1 and a huge salt.
In the last year or two several password troves have become available to all. In the past researchers didn’t have a way to report on user password selection. After a myspace phishers collected passwords leaked, researchers now had a large collection of legitimate passwords. Many of the passwords were tremendously weak and thus not comparable to the enterprise password.
When setting out to crack passwords, it is helpful to figure how how the users select the passwords. This allows the cracker to have a better chance at success.
I was hoping to take from this lecture a script to analyze a list of passwords and display the tendencies found. I would like to be able to easily run a report that says: 30% of users passwords were reveals in testing. Of those 90 percent were in the format Aaaaaa11 (A=upper, a=lower, 1=any number). I don’t see that script on his website, I’m going to check back later.
They’re hacking Our Clients, Why are we focusing only on servers; Beale
This talk had two major sections. The need for patching clients, and a poor man’s way to find clients that need patching.
In the first section Beale said that in pentesting engagements they now attempt to get to the internal network through client side attack. Often they are limited by engagement rules to the computers belonging to IT staff or security folk. Even with this set of users they are consistently able to perform attacks on the browser, mail client, Office, Adobe Reader, etc. Core Impact and Metasploit are two tools mentioned.
The bad guys moved to client side attacks years ago. Their biggest problem is managing all their owned boxes.
The question is asked, isn’t this just social engineering. There are two responses to this. No, sometimes attacks autorun without user interaction. Yes, but the human firewall is imperfect. Even the most educated users get fooled. Its still appropriate for a pentest.
Comment from the audience – Once it reaches the user, freakin game over.
The attackers only have to find one vulnerable human or one vulnerable software install.
Isn’t this a patch management problem, Beale asks rhetorically.
He says yes, but not every organization has patch management.
Also patch management, needs know about every system to patch it. It needs rights. It often doesn’t patch every product. Most people don’t have that complete an inventory of what is on their network.
To address these issues, the speaker proposed using User-Agent strings to self identify vulnerable systems. That information could be collected in HTTP proxy logs, and email servers. Vulnerable clients could be denied further access.
While you could do further things such as implement something like the Master Reconnaissance Tool to gather browser plug-ins, there is still vulnerable software that you don’t address in this way.
Another idea is to look at the metadata for recently created files on your fileserver, sharepoint, in email. Apparently you can determine the version of the software used to create the document. A vulnerable version and a recently created document equal a problem that needs to be addressed.
Since I do vuln scan all online systems, and I do have a patch management system, the second part of the talk wasn’t as interesting. It seemed like a lot of work just to catch a small number that missed the patch management and vuln scanning. I do see the usefulness in a University or other similar environment.
VOIP Hopper; Ostrom and Kindervas
This was strong talk demonstrating their new version of their voiphopper program. Most people outside that room think that a vlan is a security separator. The talk showed how easy it is to get onto the voice vlan. In IT there is also a low awareness of VOIP threats. People think, “you can’t access corporate data from an IP Phone.”
voiphopper now includes a Cisco Discovery Protocol generator making it really easy to pretend to be a VOIP phone.
Mitigation-
1. Use Cisco’s phone CDP Security provided in 12.2.36 SE. This requires a phone to have power or it will shutdown the port. (one wonders how that would work in my case where a bad blade wasn’t providing power for some ports, and I was given a brick for my phone instead of using power over ethernet).
2. MAC address filtering
3. Disable the pc port on the phone. (this is the lobby phones that should be have a pc plugged into them).
Got Citrix? Hack it!; Gupta
One audience member correctly asked for less IE vulnerabilities and more about Citrix I agree. The vulnerabilities presented all existed because Windows was not secured for the role the system was playing.
Gupta has a good point that people think putting something behind Citrix is equal to securely serving it.
We did not get to see a couple of demos because the wireless network was down during this session. I’d recommend either not relying on a unreliable medium for a presentation or have a video backup. We were left with a session cut short, and a feeling of disappointment.