[FDE] Hyperthreading not good for traditional cryptography
Ali, Saqib
docbook.xml at gmail.com
Sat Mar 3 14:19:31 MST 2007
http://www.irisa.fr/activity/new/007/branchpredictionattack004?set_language=en
No we we are not talking about Quantum Computer. We are talking about
the Intel's Hyperthreading processor that is capable of executing two
threads in the same cycle to better utilize the hardware.
Unfortunately, running two threads in parallel on the same hardware
CPU can lead to some information leakage. "One can manage to grab an
indirect view on a thread execution from a spying thread that is
executed simultaneously. This indirect information about its
execution can allow to recover critical information such an encryption
key."
Onur Acıiçmez , Çetin Kaya Koç and Jean-Pierre Seifert were able to
grab 508 bits of a 512 bit RSA Key.
Now this is NOT a traditional brute-force attack and the cracker must
be able to access the machine where some spyware has to be installed
André Seznec of "Institut de recherche en informatique et systèmes
aléatoires" (Irisa) was able to independently verify Onur's claims.
"I've tried to validate the principle. It works! Beautiful case study
by the way!" said André
More details at:
http://www.irisa.fr/activity/new/007/branchpredictionattack004?set_language=en
More information about the FDE
mailing list