Adam Boileau is the lead CANVAS developer at Immunity Inc. and fairly well known for pulling innovative new attacks out of thin air on a fairly regular basis. Boileau's best-known research went into breaking into locked Windows workstations by manipulating their Firewire interfaces.
With the recent "cold boot" attacks making headlines, Adam thought it was time to release his attack tool -- Winlockpwn -- to the public. His attack doesn't require the removal of RAM but is just as effective as the "cold boot" attacks in unlocking Vista systems running Bitlocker. It gets better -- if your target is a laptop computer that doesn't have Firewire that doesn't matter, just pop in a PCMCIA Firewire card and Windows will very helpfully install the drivers for you.
In this Q&A, Boileau discusses his Winlockpwn tool with SearchSecurity.com.au contributing editor Patrick Gray.
Adam Boileau (AB): Well it's a tool that I wrote when I was working on the Firewire stuff that I demonstrated at Ruxcon a couple of years ago. I could gain access to a target system through the memory of a remote system via Firewire. Obviously you could read memory with that and you could write memory with that and if you were sufficiently cunning then you could do relatively evil things with it. Of course being somewhat evil people we wrote a tool, myself and Brett Moore, that would allow you to unlock locked Windows machines or log in without a password or spawn an administrator shell at the login prompt without having to log in, merely by plugging in your Firewire cable and running a command.
We chose not to release it at Ruxcon because Microsoft was a little cagey about whether Firewire memory access was a real security issue or not and we didn't want to cause any real trouble. But now a couple of years have passed and I think everyone has agreed that it is a feature rather than a bug and I think it is time to release it and discuss it.
Patrick Gray: You said back in 2006 that tools that are designed to lock down Firewire ports are a little bit ineffective. Do you have a shame list?
AB: Of the tools that I tested that purported to lock down USB ports and Firewire ports some worked and some didn't. In general the ones that didn't work were the ones that allowed you to configure specific classes of devices to be permitted. So if for example you are allowed to have Firewire video cameras but not Firewire discs and obviously the secret sauce in the stuff that I demoed was the ability to lie about what you were over Firewire.
"I am a camera not a bearded Unix freak," that sort of thing?
AB: Exactly! You have literally set a descriptor flag that says "Oh hi, I am like, totally an iPod". At that point Microsoft lets you in. The older pieces of software I looked at, ones that completely disabled the Firewire, work. Ones that allowed more granular control didn't.
Getting onto those cold boot attacks, really they don't do anything that your Firewire attack doesn't do right?
AB: Well no, not really. It is another way of getting access to the memory of a machine. It is a little more crude because it is read only whereas my Firewire attack allows you to have write access to memory.
So you reset the password to A or whatever?
AB: Yes or in the case of the Windlockpwn tool it just overwrites the code branch which gets the password to say "well, any password is fine with me, that's cool". As Microsoft said, physical access is always going to win and these memory issues have been known about for along time. But I think the reason that this is interesting is that they have developed a practical, workable, usable attack that they can demonstrate to people who aren't that technical
Bitlocker has actually remained, in essence, since its release. If you are using Bitlocker with pre-boot authentication -- which means your keys aren't loaded into memory until you've entered a password or secret -- and you keep your laptop switched off or in hibernation mode when you are lugging it around, your data is actually safe, right?
AB: If you keep it switched off, yes.
Well that is the first time we can really say that isn't it?
AB: Well, yes and no.
It is the first time that full disc encryption has actually been a practical and easily deployed solution. That has really come about in the last couple of years, surely.
AB: Yes I agree with you there. Attacks against most business systems that are using full disk encryption other that these sort of attacks where the machine is on or suspended or in hibernate, attacks against off machines is really limited to double handling attacks where I can go into your office, Trojan the system to subvert pre boot configuration protections on the machine, go away again and wait for you to enter the crypto keys, come back and steal it again. That sort of double handling. There are options, but no, in terms of just stealing an off laptop that has full disk encryption we are in a much, much better place than we were even six months ago.
You must wonder what the ramifications are for the data centre because that is the one place where you are not going to use pre boot authentication.
AB: Or indeed turn your machines off.
If you get a power outage or a forced reboot or whatever you don't want to have to have someone physically in the data centre to actually type in a password. You want it to boot up to a log in screen. That is more of a worry isn't it? If there's physical access to your machines in your data centre it's pretty much game over.
AB: Yes, and I guess most people who buy commercial-grade data centre hosting assume that they have some degree of physical security because they are in a data centre. As to how much real security that buys you... some guy could still walk in with a router under his arm and an ID badge. You hear stories about people wandering through data centres and poking around finding root shells on consoles.
Well, physical attackers can boot stuff into single user mode or even rent rack space in the next rack.
AB: Exactly. If your attacker's budget goes as far as renting rack space in the same colo facility then yes you are in a pretty bad place.
The text above is a transcript taken from ITRadio's Risky Business podcast.