They Can Hear You Through The Wall. And Nothing In The Room Knows.

Picture a quiet room. The door is shut. The line is encrypted. Everything you were taught to check is green.

And somewhere outside, past the wall, someone is listening to every word.

They did not pick the lock. They did not break the encryption. They did not touch the device at all. They pointed a radio beam at it from across the street, and the ordinary electronics inside did the rest, quietly turning private words into a signal that travels back out through the wall and into their hands.

Here is the part that kept bothering me. We have spent decades building better locks. Passwords, keys, encryption, firewalls. And this walks past all of them, because it does not happen in the software where our locks live. It happens in the physical hardware, underneath. The encryption is still on. It simply has nothing to protect against, because the leak is not in the data. It is in the physics.

And the theft is not even the worst of it. The silence is. There is no alarm. No warning light. No line in a log. The conversation leaves the room and the room never knows. You could sit through the most sensitive meeting of your life while it is being emptied, and walk out believing it was safe.

For a long time the honest answer to this was to hide better. Thicker shielding. More filtering. A heavier wall. But a wall you cannot see through is also a wall that cannot tell you when someone is on the other side of it.

So we asked a different question. Not how do we build a thicker wall, but is there something the attacker does not have, and the device does?

There is. The device knows its own voice.

It is producing the signal, so it knows exactly what it sounds like from one instant to the next. The attacker, out on the street, is guessing. The device is not.

That is the whole idea behind IME-DET. It does not try to stop the leak. It listens for it. It listens to what the device is giving off and asks one simple question: is a copy of my own voice riding out on that beam? Because it knows its own voice precisely, it can pick that faint echo out of a noisy world, even when the leak is far too quiet for anything else to notice. And once it hears it, it can point to the exact frequency the attacker is using.

The part that stops it crying wolf is just as important. A simple detector measures how much radio noise is in the air, and the air is always full of it, phones, wifi, everything. So it either alarms all day or misses the real thing. IME-DET does not measure how loud the room is. It checks whether the sound is specifically the device's own voice. It reads the letter instead of weighing the envelope. Harmless noise, however loud, is not your words. Only the real theft carries your words.

So for the first time, an attack that was designed to be invisible has something looking straight at it. Not a heavier wall. A pair of ears that knows your voice.

I want to be honest about where this is. IME-DET is early. It works in simulation, where it catches the quiet theft that ordinary detectors sleep through and tells a real leak from ordinary noise. It has not yet stood on a bench against real hardware, and that is the next test, the one that decides everything. This is the beginning of the story, not the end of it.

But the shift underneath it is the one we keep coming back to at Physivitis. The problem was never that our software locks were weak. It was that the leak lived one floor down, in the physics, where the locks do not reach. So that is where we went to meet it. Because the walls that hold up the digital world are not made of code. They are made of physics. And that is the floor we build on.