Skip to content

Observable cryptographic hardware

Observable cryptographic hardware reports on its own state — entropy health, key operations performed, algorithm and parameter set in use, and firmware version — as authenticated telemetry an operations team and an auditor can consume. It extends the LOVE argument from the entropy source to the whole cryptographic boundary.

The generalisation

The entropy argument is that a source which measures and reports its own health can be trusted currently rather than historically. Nothing about that reasoning is specific to entropy.

A cryptographic boundary that reports which algorithms it used, with which parameters, on which firmware, and what its entropy health was at the time, is answerable in the same way — and answerable is what an auditor is actually asking for.

What such hardware would report

  • Entropy source health: the model's parameters as values, and the resulting status.
  • Algorithm and parameter set actually in use, which is frequently not what the configuration claims.
  • Firmware version and measured boot state, so what is running is not in question.
  • Operation counts and rates, enough to detect anomalous use without recording the operations themselves.
  • Key lifecycle events — generation, rotation, destruction — without exposing the material.

Authentication is the hard part

Telemetry an attacker can forge is worse than none, because it manufactures false confidence. Health reports have to be signed by the same root of trust whose state they describe.

That creates a bootstrapping problem worth being explicit about: the report is only as trustworthy as the key signing it, and that key came from the entropy source being reported on. The chain closes on itself, which is an argument for the source being inside the boundary that signs — the root-of-trust design.

Where the evidence has to go

Into the systems an organisation already uses: monitoring, SIEM, and whatever retains records for audit. Telemetry in a vendor-specific console is not evidence during an incident, and it is not there in three years when the question is about a key issued today.

Retention is the part usually forgotten and the part that decides whether provenance questions are answerable at all — which is an operations decision, not a product feature.

Why this closes the loop

Assurance means a model, measurement and inspectable evidence. Applied to everything the hardware does rather than to the entropy source alone, it is what makes a boundary answerable instead of merely certified.