[boredgamer2]

> TEMPEST [1] (or Van Eck Phreaking) is a technique to eavesdrop video monitors by receiving the electromagnetic signal emitted by the VGA/HDMI cable and connectors

I've always thought of TEMPEST as magic and incredibly cool. I get how this works for video, but why is it that infosec recommends for EM shielding of a datacenter? Wouldn't there be so much EM noise from a DC that the data you could pull off be .. noisy? lossy?

[pjc50]

It's difficult, but it gets a lot easier if you can send requests in. If you can force some kind of deterministic (not salted) crypto calculation to happen with a particular piece of key material, over and over again, you can use the same correlation technique that works for extracting GPS signals from hundreds of dB below the noise floor.

People have also done this with analysis of the power consumption of crypto hardware. I believe this was used against Trezor devices.

[ohazi]

> GPS signals from hundreds of dB below the noise floor

Nitpick: GPS is more like 20 dB below the noise floor

[1] https://sdrgps.blogspot.com/2016/02/find-signal-in-noise.htm...

But yeah, theoretically you can cook up as much process gain as you need, it just might take a while.

[madengr]

My old boss always said "There isn't 200 dB in the entire universe". Of course if you compare the diameter of the universe to a quark, it's 10,453 dBmeters, so I guess there is.

[jandrese]

200dB is the one way freespace path loss to Geo orbit.

Back when I learned this I was shocked that satellite communications work at all.

I can only imagine how much loss you have trying to hear communications from the Voyager probes now.

[drmpeg]

318 dB for Voyager 1 at X-band.

[boredgamer2]

> Back when I learned this I was shocked

Care to explain? I'm not following..

[jandrese]

Every 3dB is roughly a halving of the power level. It's hard to wrap a human brain around the orders of magnitude difference, the numbers are so far outside of anything we can encounter in the human scale.

Still, I will try to make an analogy. Start with the smallest transistor we make today, (7nm) and scale it up by the equivalent of 200dB and it would be 258 million km across, or about the same as the diameter of the Earth's orbit around the sun.

[ohazi]

That's a pretty good rule of thumb until you get into some really weird stuff.

The link budget for bouncing radio signals off the moon (EME) is something like 260 dB. A while ago I had a chance to see the ground station at NASA Goldstone where they manage the radio links to Mars and the outer planets. There was a terminal open with ~single digit bits / second coming in from one of the Voyager probes. Having a 70 meter dish and a cryogenic receiver helps, but the link budget there has got to be truly staggering...

[appleflaxen]

what does that mean? can you dumb it down at all for a non-signals person?

[madengr]

My old boss worked on the Apollo mission comms, specifically the large dish antennas on the ship. It was just his way of joking about large numbers.

10 dB (deci-Bells) is an order of magnitude ratio; 10 dB = 10Log(10). For non ratios you tack on units, such as 30 dBHz = 10 Log(1 kHz). It just a way of expressing large values in engineering, and you can add the dB instead of multiplying in linear domain. You begin to think in dB after doing it for years.

The path losses stated in the replies are good examples when the rule is broken. The path loss is 22 dB + 20*Log(distance/wavelength). My universe/quark is just a joke of the most extreme ratio I can think of; I’m sure there are others larger.

The national debt ought to be in dB$.