Stray electronic-magnetic leaks used to harvest PC crypto keys Israeli security researchers have been able to extract encryption keys from a nearby computer by analysing stray electromagnetic radiation. The attack by computer scientists from Tel Aviv University shows that TEMPEST-style side channel attacks are no longer just the preserve of Mission Impossible and three-letter spy agencies …
So am I understanding correctly that to pull off this attack, they had to get the target to decrypt a carefully chosen cyphertext first, several times, and THEN they could decrypt whatever it was?
It seems a little impractical for anything beyond nation states who can already do this, or is that just me?
How practical an attack is this? If I'm reading it right the eavesdroppers would need to give a know file to their target and ask them to encrypt it using the same key as they use for their other files, then obligingly decrypt it several times so they can pick up the key, then the attackers could access the other files.
Surely this is trivial to jam? My office is full of various PCs and electronics - often without cases. Are they going to pick out that one signal of the encryption being decoded? And what happens when I turn on my Plasma TV - that blasts MW and LW radio in a wide area.
This seems a little too much of "proving a theory" but not being practical in real world use.
Radio antennae can be made highly directional and highly frequency-selective (like a Yagi), allowing you to zero in on your target with enough precision to penetrate most of the noise.
As for getting enough iterations, just use a commonly-used encrypted target and hit the sniffer each time it's being decrypted (you can hide the device in a box in a closet, maybe).
its how the Enigma crack worked.......because the Germans tended to send routine identical signals at similar times of day it was possible to presume what a message held and then brute force it
for instance every weather message would have a preface indicating time / location, once thats cracked the rest is open - especially as weather reports follow a predictable format
in this case, if the PC next door was also send messages in a predictable format (stock reports, bank details, sales reports) then cracking it should be relatively trivial.
This is going to sound daft, but back when I worked for Time Computers the meeting rooms had windows with a metal coating, secondary glazing and fixed blinds, all in an attempt to stop electronic or lazer eavesdropping.
What the brothers were worried about is anyone's guess......the only real secrets were the true ownership of all the companies in the fraud......,and just who in Dubai got all the money
(thinks in background.....Bin Ladin, Bin Ladin). Totally no proof of course
Point being TEMPEST is such onerous that only the most important of computers are given the treatment. Now, these researchers are saying it can be done ANYWHERE, with inexpensive, not-too-hard-to-conceal equipment, meaning practically nothing is safe. Imagine the effect on business budgets if every single piece of electronics in their firms had to be given TEMPEST-level hardening.
But what I found interesting about the presentation on their website is that, in the first picture's caption, they list items from left to right. Native Israelis (and the rest of the Middle East) tend to enumerate things from right to left.
Most confusingly, the "first" item in a shelf may be the rightmost or the leftmost one from the point of view of the same person, purely depending on whether they are speaking Hebrew* or English or Russian.
* I haven't seen this with Arabic speakers, but I presume the same phenomenon would hold for a native bilingual.
Soooo, if they (currently) require multiple runs to capture/identify the signals - presumably any machine running "other code" at the same time (as most machines do?) might well create obfuscation? Also, why not simply build in random calls during the decrypt, for optional "secure decryption" - every X operations, head -Y /dev/random into /dev/null (or /tmp/file), and use a decent RNG to generate X & Y. It'd slow the process down, but theoretically create enough chaff to hide things, and wouldn't be the same on any replay?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
