TrojPix: air-gap exfil via video-cable RF emanation

The cable running from a graphics card to a monitor is not supposed to be a transmitter. The shielding is meant to hold the video signal inside the physical link. It doesn’t, and it never has — every DVI, HDMI, or DisplayPort cable radiates a faint radio signature that mirrors the pixel data it is carrying. Researchers at Shandong University have shown a new way to shape that leakage into a covert exfiltration channel out of an air-gapped machine, summarized by The Hacker News on July 6. They call the technique TrojPix.

The mechanic, at the highest level, is this. Software on the source machine writes specific pixel patterns to the display buffer. The patterns are chosen so that a human looking at the screen sees a normal image — the changes sit below the threshold of visual perception — while the electromagnetic emissions from the video cable, keyed to those pixel patterns, encode a distinct bit stream. A receiver with an appropriate antenna in the vicinity picks that stream off the air and decodes the exfiltrated data.

This is a member of the TEMPEST family of side-channel attacks, which have been in the public literature for decades. The novelty in TrojPix is on the encoding and speed side, not on the physics.

The precondition nobody should skip past

The Hacker News summary contains the sentence that determines whether this matters to you or not: TrojPix works only once malware is already on the target machine. The technique moves data out of an air-gapped host. It does not move malware in. Something else — a USB, a supply-chain firmware implant, a compromised install medium, a briefly-connected update laptop — has to have put the code on the machine first.

If your air-gap is real and no one has ever put anything on the isolated host, TrojPix is not the day’s problem. If your air-gap is the polite kind — enforced by policy and periodic sneakernet updates — then this is one more reason the policy is not the same thing as the physics.

What this changes about the threat model, and what it doesn’t

It doesn’t add a new intrusion vector. It reinforces something operators of genuinely sensitive isolated networks were already meant to be treating as design constraint: EM containment is part of the boundary. In classified environments this has a name and a specification set — NATO SDIP-27, NSTISSAM TEMPEST/1-92 in the older US lineage — and expensive facility construction to back it up. TrojPix does not break those environments; the point of the containment budget is that the emissions do not leave the room to begin with.

Where this lands practically is in the tier below classified: ICS/OT segmentation zones, research labs, financial audit systems, hospital records enclaves. These are the setups where an “air-gapped” workstation is a normal workstation in a normal room, with a normal display cable that runs across a normal desk to a normal monitor, and the receiver hypothesized in the TrojPix threat model can sit in the next room over or in a parked vehicle.

The specific detail worth carrying away

The single defensive detail that is not obvious from the news summary and that legacy TEMPEST guidance has consistently emphasized: shielded, properly-terminated video cables of known provenance are not the same object as a bargain cable from a bin. If the isolation matters, the cable is part of the boundary — replace commodity cables with rated shielded assemblies, keep the display physically inside the same EM-controlled space as the host, and confirm that any exfiltration path from the host to the display does not pass through a wall you cannot survey. The rest of the mitigation stack — no untrusted media, no unattended access, no bridging laptops — is the same as it has been for thirty years.