cold boot attacks on disk encryption

[Ivan Krstić]

On Feb 21, 2008, at 6:40 PM, Ali, Saqib wrote:
> i think in most cases tamper-resistant is sufficient


Er, what do TPMs have to do with this at all? TPMs are not tamper- 
proof hardware FDE devices. They're somewhat tamper-proof (in  
practice, I wouldn't depend on it) non-volatile storage for small  
amounts of sensitive data, such as encryption keys. But as long as  
it's software that's driving your FD encryption, you need to have your  
keys in RAM.

So, either:

* The TPM is used in 'basic' mode, where its only purpose is to
   provide a measure of tamper-resistance to the boot path, and as
   long as no boot-time tampering is detected, the FDE key will be
   loaded into RAM automatically,

or,

* The TPM requires explicit authentication (e.g. by password or
   smart card) before releasing the key, in which case a successful
   authentication will load the FDE key in RAM.

If the machine is running and the FDE in use -- which is the entire  
premise behind this attack -- both TPM use cases are just as  
vulnerable. TPMs are a red herring in this discussion, unless the FDE  
was actually offloading the crypto operations to it. This is not a  
supported mode of operation for any widely-deployed FDE system that  
I'm familiar with.

So, is anyone else as amused as I am that Apple can release an EFI  
firmware update to zeroize MacBook Air memory at boot-time, turning  
the heretofore widely-decried inability to upgrade that laptop's RAM  
-- due to the chips being soldered to the motherboard -- into an  
advantage, and making the Air the laptop of choice for discriminating,  
fashion-aware, security-conscious professionals the world over?

--
Ivan Krstić <krstic at solarsail.hcs.harvard.edu> | http://radian.org

---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at metzdowd.com