Maybe I'm missing something, but this reads like a complicated way to say "We made an IR diode that gets cold as well as hot."

While this is equivalent with "gets cold as well as hot", there is a critical difference.

Modulating the infrared emission by cooling and heating a body is slow, so the transmission rate is low and it is also easy to detect, because any infrared detector will show pulsed infrared light.

The whole point of the article is that they have found a method for modulating the infrared emission that is much faster than cooling and heating, so because the modulation frequency is so high any normal infrared detector will not see anything, it would just detect the normal infrared emission that corresponds with the ambient temperature.

They exploit a phenomenon that exists in infrared LEDs made for a low frequency (high wavelength), which when biased forward emit infrared light, like any LED, but when biased backward the reverse happens, i.e. their infrared emission is lower than it should be for a black body at ambient temperature, because a part of the thermally emitted photons are reabsorbed by the semiconductor, generating electron-hole pairs that are separated by the electric field, being thus prevented to recombine and emit again a photon.

Because the increases and decreases in infrared luminosity are done by changing the bias voltage of a LED, they can be done orders of magnitude faster than by cooling and heating.

I do not know whether this proposed application in steganography would ever be worthwhile, but this is certainly a very cool development.

Or you can call it encryption along different axis. Much like extracting GPS signals from below thermal floor level - you can do it if you 1) know it's there, and 2) know exactly how to key in. It's impressive as heck, but you can always rephrase it in terms of information theory in ways that makes it sound like slightly different shade of mundane.

I don't believe you're missing anything. This is just stegenography with a possibly new covert channel, right? Apparently the secret depends on advisaries not noticing the special hardware deployed on each end. Would using spread sprectum techniques would work just as well?

I think the reason the negative luminance is potentially important for secrecy is that it means the average of the signal you’re transmitting is zero, making it indistinguishable from noise.

Yeah, but saying that doesn't get the military to give you money.

Yep.