Entropy for key generation, nonces and signatures
Cryptographic operations consume entropy at four main points: long-term key generation, ephemeral key exchange, nonces and initialisation vectors, and signature randomness. Each fails differently when the supply is weak, and signature nonces fail worst — a repeated ECDSA nonce exposes the private key outright.
Long-term key generation
The highest-consequence use. An RSA or elliptic-curve key generated from a weak source is predictable for its entire life, and its life may be decades. Compromise is typically discovered by an outside party long after the fact, and there is no way to detect it from the key itself.
This is where provenance would matter most and where almost nothing records it.
Ephemeral key exchange
Every TLS handshake generates ephemeral key material. Weak randomness here compromises the session rather than the long-term key, which sounds better and is still a full compromise of that traffic.
The relevant property is volume: this is the consumer that drives reseeding rates, and therefore the one that turns entropy from a design concern into a capacity calculation.
Nonces and initialisation vectors
Many modes fail catastrophically rather than gracefully on nonce reuse. GCM loses authentication guarantees entirely if a nonce repeats under the same key. CTR mode leaks the XOR of plaintexts.
The requirement is uniqueness rather than unpredictability in some modes and both in others, which is a distinction implementations get wrong regularly.
Signature randomness
The sharpest failure in cryptography. ECDSA requires a fresh secret nonce per signature; two signatures under the same key with the same nonce allow direct algebraic recovery of the private key. Not weakening — recovery, from two signatures.
This is why deterministic nonce generation (RFC 6979) exists, and it is the mechanism behind the Android wallet compromises.
What changes with post-quantum algorithms
Nothing structural. ML-KEM and ML-DSA consume randomness for key generation and for operations exactly as their predecessors did, and a weak source undermines them identically. The dependency is the same one, in new algorithms.
Keep reading
- Consumption and audit
What auditors ask about your entropy source, and how to answer
Auditors ask four things about an entropy source: where randomness comes from, what evidence supports the entropy claim, what runs in service to detect degradation, and what records exist for keys already issued. The first three usually have answers; the fourth usually does not.
- Consumption and audit
Quside's Entropy Core through AWS
Quside develops new-generation quantum technologies for secure connectivity and efficient computation.
- Consumption and audit
Quside's Entropy Core through Equinix Fabric TM
Quside develops new-generation quantum technologies for secure connectivity and efficient computation.