Loading video...

Video Failed to Load

Go Home

Introducing deep-TEMPEST: a deep learning method that recovers great quality images from unintentional electromagnetic emanations of HDMI. Great work (in progress) by E. Martinez, S. Fernandez and G. Varela 💪💪 (co-mentored with Pablo Musé). Expect more news in the next weeks.

906,490 views • 3 years ago •via X (Twitter)

11 Comments

Federico 'Larroca' La Rocca's profile picture
Federico 'Larroca' La Rocca2 years ago

Since you're here, you may as well join our CTF starting next weekend:

Leandro Tami's profile picture
Leandro Tami3 years ago

@muse_pablo Scary. Besides espionage what applications could this have? I mean it's great to know this is actually physically possible but I think what could come out of this in the long run. A better HDMI that is less vulnerable to this perhaps?

Federico 'Larroca' La Rocca's profile picture
Federico 'Larroca' La Rocca3 years ago

@muse_pablo In our case, as students and teachers in a University, we're learning a lot (signal processing, programming, ML, radio and electronics). That's our main goal (besides, as you mentioned, generating awareness on the possibility).

JuanP's profile picture
JuanP3 years ago

@muse_pablo Una clara muestra de el porqué del uso de jaulas de Faraday y múltiples sistemas de filtrado ( red eléctrica) en salas de ordenadores donde se examina material clasificado.

All the Things's profile picture
All the Things3 years ago

@muse_pablo My understanding is that this is an algorithm that can basically “read” what is on your computer screen wirelessly. It does this by taking the electromagnetic field that’s around your HDMI and decodes it back into an image on whoever’s computer is running this algo.

Federico 'Larroca' La Rocca's profile picture
Federico 'Larroca' La Rocca3 years ago

@muse_pablo That's right! The challenge is coming up with a good model of the HDMI signal, both to decide at what frequency one should "tune" to, and to then recover the image (using "traditional" methods as in gr-tempest or TempestSDR, or by using more sophisticated ones as here).

James Morris 👾's profile picture
James Morris 👾3 years ago

@muse_pablo Amazing. I wonder if you could increase your S/N with a filter & LNA around the emanation frequency.

Federico 'Larroca' La Rocca's profile picture
Federico 'Larroca' La Rocca3 years ago

@muse_pablo Yep. And a good antenna would help too.

Miloš Mićić's profile picture
Miloš Mićić3 years ago

@muse_pablo It be would nice to see this setup with directional antenna and few meters of gap. Great work!

Umat Šumak's profile picture
Umat Šumak3 years ago

@muse_pablo Are you working with leaks from the cable or from the device sending out the signal? If cable, how much does it depend on the "quality" of the HDMI cable?

Federico 'Larroca' La Rocca's profile picture
Federico 'Larroca' La Rocca3 years ago

@muse_pablo Both cable and connectors. In my experience, cheap connectors are adapters are huge emitters! See for instance for an example of a cheap connector in VGA.

Related Videos