Comment by all | Hacker News Reader

Comment by all | Hacker News Reader

Before you dive too far into full-disk encryption, you might want to contemplate the consequences of this research:

http://news.cnet.com/8301-13578_3-9876060-38.html

The paper referenced is available at:

http://citp.princeton.edu/pub/coldboot.pdf

From that paper, the most salient sentence for this discussion is:

"On all of our sample DRAMs, the decay rates were low enough that an attacker who cut power for 60 seconds would recover 99.9% of bits correctly."

The long and short is that your password has to be in RAM at some point in order to be matched against what is typed in. RAM decays slowly after a machine has been switched off. As the research shows, one can use a widely-available can of air to make that decay happen even slower. In the process, one can pull the value out of RAM.

Note that this is a hardware problem that no software on the market has been able to address. To do so, you would have to be able to validate x against y without having the value of y ever enter RAM.

Consequently, a reasonably good tech thief can read the password out of RAM. For the not-so-good tech thief, using passwords and social engineering tricks like being nice to someone and then asking them to watch your stuff for a moment will probably do the trick. I tend to see the various implementations of full-disk encryption as more for peace of mind and mitigation of liability than real security.

Remanence isn't a realistic attack in his threat model; attackers have mere minutes to get the RAM out of his system, cool it, and siphon the data off. If he's worried about losing the Plans to the Empire when his ship is captured, sure. But if he leaves his bag in the back of a cab, he'll be fine.

The notion that any "reasonably good tech thief" could pull this off is also hard to take seriously. So far as I know, there are zero (0) remanence key extractor tools in circulation. Highly skilled attackers --- many apparently in the employ of foreign governments --- have a hard enough time writing tools to extract keys from live memory.

The threat is not against the Plans to the Empire but the password for the encryption system. That, as the video shows, can be done and without any particular haste. As the article suggests, freezing the RAM lengthens the decay process. The method is likely to have been known by three-letter agencies for some time now. The good guys are seldom, if ever, ahead of the bad guys, so it seems likely that the latter know it, too.

But, for the sake of argument, let's say that they don't crack it on the first go. So what? The password has to be in RAM every time the machine starts. AFAIK (and I would be pleased to be wrong about this), nobody shreds data on the hard drive because the user gets their password wrong. So, said bad guys can try to their heart's content.

As far as I'm concerned, it is a major hole in any argument for full-disk encryption.

I think you're misunderstanding both the attack and full-disk encryption. The decryption key is not in RAM every time the machine starts; you have to enter material to derive the key at bootup. And the attacker does in fact have minutes to recover whatever's in RAM once the machine shuts down --- which means that if you shut your machine down before leaving the office (which you have to do), remanence simply isn't a threat at all.

Everyone "shreds" their hard disk if they lose the password. That is the point.

Also, the Plans are to the Death Star, not the whole Empire.

Rarely does one shutdown a laptop anymore. Usually people just close the lid and the computer is put to sleep. In sleep state the contents of RAM is preserved AFAIK and should give the attacker more time to think their plan through.

The rules are different when you use full-disk encryption. You shut your machine down. PGP WDE actually prevents "sleep" from working, but regardless, you still need to be vigilant.

PGP WDE doesn't prevent sleep from working on my MacBook. It disables hibernation, but that's not a security choice. It's just a limitation of the implementation.

What you say is the right policy to maximize security, but PGP WDE doesn't enforce it.

You're right. Sorry, I misspoke.

The OP is worried about thieves, not three-letter-agencies.

True, but the point of the article is that both now have this level of know-how. So all bets are again off once physical security is compromised - regardless of whether its a farmboy from Langley or Nebraska.

You're looking at a different threat model. Hard drive encryption isn't there to stop the government, the police, or the Girl Scouts.

It's there so that when the Pink Panther walks away with your laptop, he doesn't get instant access to to all of your data. If Mr. Laptop Thief wants your home directory, he's going to need to do a hell of a lot more work than reinstalling the OS and throwing the machine up on eBay.

If your drive isn't encrypted, I'll wager that a laptop thief would take a few minutes to poke around and look for something interesting, like that file 'Corporate Strategy 2010.doc', or "My Affair With The Underaged Intern, Oh God I Hope This Doesn't Get Leaked Onto The Pirate Bay.avi".

isn't a realistic attack in his threat model

Under different circumstances your solution would be 100% correct, however, in this situation we are worried about "everyday" type criminals. These are the guys/girls that find a laptop in an unlocked locker and snatch it; small window of opportunity. In your analogy the description of the criminals is much more sophisticated; Oceans 11 types.

Stealing RAM can potentially tip someone off to a threat, a better solution IMO is the Stone-Cold Bootkit[1]. The bootkit can hook into Windows in order to gain unrestricted access. It does this by writing to the MBR unrestricted; haven't we seen this before[2]? Once installed it will log your decryption password and do what ever with it.

A BIOS password(windows) and HHD encryption will be enough to keep the NSA out for a while, and a normal criminal out indefinitely.

It would be nice to have a mobo with some type of secure memory encryption programs could write to- digitally signed. Of course that could be owned too, but it's a start. In the words of Bruce Schneier, "[sic] it is very difficult to secure data when the attacker has physical control of the machine the data is stored on."[3]

Now I am thinking about the /boot partition in linux; where GRUB is stored.

[1] http://www.stoned-vienna.com/

[2] http://news.ycombinator.com/item?id=1643451

[3] http://www.schneier.com/blog/archives/2008/02/cold_boot_atta...

[4] PSA: http://zedomax.com/blog/2009/02/23/linux-boot-hack-how-to-pa...

You've cited Schneier citing the same Felton paper that 'all cited. This research, while interesting, does not mean what 'all think it means. It is simply not the case that an attacker who captures your powered-down machine is going to be able to use remanence attacks to recover your encryption key.

The real-world attack Felton's paper implicates is this: an attacker captures your powered-on laptop, because you left it sleeping in your bag and your full-disk encryption software can't reliably wipe keys when it wakes from sleep. The attacker restarts the system and uses remanence to dd the previous RAM contents into a drive for analysis.

This very well may be an attack that government agencies are prepared to use against laptops, but since it's not going to work against anyone who knows how full-disk encryption works, it's simply not going to be in the arsenal of someone who steals your bag out of a bar.

There are always way to hack a computer: but we're just asked to take reasonable steps toward making it harder.

Yes, but that's exactly why it's important to realize this only gives you some protection. The OP asked for his personal needs, not because of some company policy, which means he/she should know whether it really is effective.

The poster mentions they wish to defend against theft and lost.

The attacks you cite are more applicable if he were worried about intelligence agencies intercepting his data. And with rumored incestuous relationships with the corporate world, that might not be a bad idea, if he has valuable IP.

Reply on news.ycombinator.com