Hardware
Academia keeps improving zero-knowledge proofs of photo and video editing: Eva, VerITAS, zk-Cinema. They all assume the original video carries a trustworthy camera signature over the right representation of the pixels, but they do not address how a camera produces that signature securely. The camera-side hardware for provable photos and videos is yet to be built.
Full spec
Architecture, register-level interface, configuration lockdown, sequence diagrams, and negative tests, written for an SoC or firmware team. Read it on GitHub.
Loading sequence diagram…
None of this is exotic silicon. TPMs already hold digests that software cannot overwrite and sign them on request; phone secure elements already sign with keys that never leave the chip; hardware encoders already compute the pixel samples the binding engine would consume. The design is deliberately general: different proof systems (hash binding for Eva, commitment binding for zk-Cinema) are different locked settings of the same block.
The hardware side is a small, well-scoped integration: locked pixel tap, streaming binding engine, and a narrow secure-element attest. Pilots can start small (firmware or FPGA), and the interface is meant to become an open spec that early partners shape.
If that overlaps your roadmap, get in touch. Critical feedback on the spec is welcome too.