Designing a root of trust around local entropy
Designing a root of trust around a local entropy source means the source sits inside the same boundary as key storage and verification logic, and its health is measured by that boundary rather than asserted from outside. The design decisions are where the source sits, what crosses interfaces, and what the system does when the source degrades.
The boundary
Everything the root of trust protects, and the entropy source that seeds it, inside one boundary. Key material generated inside is never exported. Raw entropy never crosses an external interface, because raw output is the most informative thing about a source and it should not be observable outside.
What crosses outward is narrow: public keys, signatures, attestation statements, and health status.
Health status is an interface, not a log line
A root of trust that measures its own entropy source has to expose the result, or the measurement is private and useless. The interface is small — a status, the underlying parameter values for trend analysis, and an alarm — but it has to exist by design.
It also has to be trustworthy, which means signed by the same root of trust reporting it. Unauthenticated health telemetry is a channel an attacker can forge to hide a degraded source.
What happens when the source degrades
The decision worth making early. A root of trust whose entropy source has failed its health checks should refuse to generate new key material, and continue to use existing keys for verification and signing.
That split matters: refusing everything turns an entropy problem into a total device outage, while refusing nothing means generating keys from a source known to be outside its model. Refusing generation only is the proportionate response.
First boot, again
The design has to work when the device has no history. That is the whole point of a local physical source: it produces entropy on the first power-up on the production line, when accumulated system state does not exist. Provisioning is the moment this pays for itself.
What this enables
A device that can state, for a key it holds, that the key was generated inside its boundary from a source whose health measurements were within range at that moment. Almost no deployed device can say that today, and it is the device-level form of key provenance.
Keep reading
- Secure silicon
Entropy IP licensing, integrating a quantum source into your own CMOS
Licensing an entropy source means integrating the source and its supporting logic into your own silicon rather than buying a component. What transfers is the design, the entropy model and the characterisation methodology; what does not transfer is a certificate, because the evaluated device becomes yours.
- Secure silicon
Photonic and electronic co-integration, how the entropy engine gets small
Co-integration puts photonic components — laser, waveguides, detector — alongside the electronics that drive and digitise them, in one package or one process. It is what takes a quantum entropy source from an optical assembly to something that can be placed inside a chip design.
- Secure silicon
What is a security chiplet, four functions in one boundary
A security chiplet places entropy generation, runtime verification, root of trust with key storage, and cryptographic execution inside one boundary, as a die that can be integrated into a larger package. It turns a set of components a customer assembles into a block they place.