The Analog Hole: Why Signatures Aren't Enough
What is the Analog Hole?
Imagine you have a smartphone with perfect cryptographic attestation. Every video it captures gets a signature from a tamper-resistant chip, proving the data came from that device's physical sensors.
Now point that phone at a 4K OLED monitor displaying a deepfake. Press record.
The resulting video has a valid cryptographic signature. The attestation is correct: the data genuinely came from the device's camera sensor. But what the camera saw was a screen, not a real scene.
This is the analog hole: the gap between proving device identity and proving scene authenticity.
Why This Matters
Content authenticity initiatives like C2PA are valuable. They establish provenance and detect post-capture tampering. But they don't solve the analog hole problem.
An attacker who records a deepfake on a legitimate device produces content with legitimate credentials. The signature is real. The attestation chain is valid. The content is fake.
Closing the Gap
To close the analog hole, we need evidence that the captured scene was real, not a recording of a recording. This requires detecting the physical signatures that distinguish:
- Real 3D scenes from flat 2D displays
- Broadband natural light from narrow-band display emission
- Continuous time from discrete refresh cycles
Each of these domains offers potential signals. None is perfect. The challenge is combining them into something useful.
Research Questions
This is where Axiom's research sits: exploring which physical constraints are actually detectable with commodity smartphone sensors, and under what conditions.
We don't have complete answers yet. But we think the questions are worth asking.
