QRNG for HSMs and PKI
An HSM generates keys that protect everything else, so the entropy source inside it is the root of the whole hierarchy. What matters for an HSM or a certificate authority is not only that the source is physical, but that its entropy claim is evidenced within the module boundary and can be demonstrated during evaluation and audit.
Why the entropy source is the root
An HSM's value is that a key generated inside it never leaves in the clear. That guarantee is about confidentiality, and it says nothing about whether the key was unpredictable in the first place. A perfectly protected key drawn from a weak source is a weak key that is very well looked after.
For a certificate authority this compounds. A CA root key protects every certificate under it for the life of the hierarchy, which may be decades. It is generated once, in a ceremony, and the entropy available at that moment is the entropy that key will ever have.
What the module boundary changes
FIPS 140-3 and Common Criteria evaluations are scoped to a defined cryptographic boundary, and entropy has to be accounted for within it. An entropy source outside the boundary raises questions an evaluator will ask: how the entropy crosses, what protects it in transit, and whether the claim about it can be made at all inside the module's scope.
This is a large part of why entropy for an HSM is generated locally. It is not only a security preference; it is what makes the evidence tractable within the scope of the evaluation.
What an HSM vendor has to be able to show
- The entropy source's model and its min-entropy estimate, conditional on that model.
- Which SP 800-90B or AIS 31 analysis applies, at which version, and to which configuration.
- What health testing runs in service, and what it detects — whether the source is characterised or observed.
- How the module behaves when the source fails: refusing to generate keys is the correct behaviour, and it should be demonstrable.
For a CA operating at scale
Issuance volume changes the shape of the problem. A CA signing at high rate is reseeding continuously, and the entropy budget becomes an operational quantity rather than a one-time ceremony concern — which is where sizing matters.
The questions auditors ask about all of this are the subject of the entropy audit page, and where entropy sits inside a hardware root of trust is the hardware pillar.
Keep reading
- Deployment and integration
QRNG form factors, from PCIe card to silicon IP
Quside supplies entropy sources as PCIe cards, FMC modules, embedded modules and licensable silicon IP. The choice is governed by where entropy needs to be generated relative to what consumes it, and by how many units you are building.
- Deployment and integration
Integrating a QRNG: C, Python, Rust and the Linux entropy path
On Linux a hardware entropy source is normally exposed through the kernel random subsystem, so applications keep calling getrandom() and inherit the better seed without changing. Direct access through a device interface or a library is available where an application needs entropy that has not passed through the kernel pool.
- Deployment and integration
QRNG throughput explained, why 1 Gbit/s matters and when it does not
Entropy throughput matters only in relation to how fast a system consumes it. The sizing calculation is reseed interval multiplied by concurrent DRBG instances, plus any bulk consumption; for most workloads that lands far below a gigabit per second, and for a few it lands above it.