While we don’t think this qualifies as a “fail”, it’s certainly not a triumph. But that’s what happens when you notice something funny and start to investigate: if you’re lucky, it ends with “Eureka!”, but most of the time it’s just “oh”. Still, it’s good to record the “ohs”.
Gökberk [gkbrk] Yaltıraklı was staying in a hotel long enough that he got bored and started snooping around the network, like you do. Breaking out Wireshark, he noticed a lot of UDP traffic on a nonstandard port, so he thought he’d have a look.
Continue reading “Secret Listening to Elevator Music”
TEMPEST is the covername used by the NSA and other agencies to talk about emissions from computing machinery that can divulge what the equipment is processing. We’ve covered a few projects in the past that specifically intercept EM radiation. TEMPEST for Eliza can transmit via AM using a CRT monitor, and just last Fall a group showed how to monitor USB keyboards remotely. Through the Freedom of Information Act, an interesting article from 1972 has been released. TEMPEST: A Signal Problem (PDF) covers the early history of how this phenomenon was discovered. Uncovered by Bell Labs in WWII, it affected a piece of encryption gear they were supplying to the military. The plaintext could be read over that air and also by monitoring spikes on the powerlines. Their new, heavily shielded and line filtered version of the device was rejected by the military who simply told commanders to monitor a 100 feet around their post to prevent eavesdropping. It’s an interesting read and also covers acoustic monitoring. This is just the US history of TEMPEST though, but from the anecdotes it sounds like their enemies were not just keeping pace but were also better informed.
A team from Johns Hopkins University has discovered a way to eavesdrop on encrypted voice streams. Voice data like the kind used by Skype for its VoIP service sends encrypted packets of varying sizes for different sounds. The team learned that by simply measureing the size of the packets, they could determine what was being said with a high rate of accuracy. VoIP providers often use a variable bit rate to use bandwidth more efficiently, but it is this compression that makes audio streams vulnerable to eavesdropping.
The team’s software is still in its early stages of development, yet incapable of parsing entire conversations. It is capable, though, of finding pre-determined keywords and inferring common phrases bases on the words it detects. It also has a higher rate of accuracy in identifying long complicated words than short ones. The team’s goal was not to eavesdrop, but to expose the vulnerability; team member [Charles Wright] notes, “we hope we have caught this threat before it becomes too serious.”
[via Schneier on Security]