In a 2012 internet-wide survey, researchers recovered RSA private keys for 0.50% of TLS hosts (opens in new tab). Nobody broke the math. The devices had generated their keys with too little randomness.
Starting January 1, 2027, NSA's CNSA 2.0 algorithms are required in all new acquisitions for National Security Systems, according to NSA's CNSA 2.0 FAQ (opens in new tab). The suite adds ML-KEM-1024 and ML-DSA-87, but none of those algorithms creates its own randomness, and NSA says commercial random number generators in these systems must generally satisfy NIST SP 800-90 A, B, and C.
If you are planning the move from RSA and elliptic curves to ML-KEM and ML-DSA, this post covers the part that is easy to miss: the entropy those new algorithms consume, and where it fits before the deadline.
What Is the CNSA 2.0 Deadline in January 2027?
CNSA 2.0, the Commercial National Security Algorithm Suite 2.0, is NSA's list of approved algorithms for protecting National Security Systems (NSS) against a future quantum computer. NSS are the U.S. government systems whose cryptographic rules NSA sets. The date to circle is January 1, 2027. NSA's CNSA 2.0 FAQ (opens in new tab) says CNSSP 15 requires CNSA 2.0 algorithms in all new NSS acquisitions by that date.
The algorithms are already standardized. Here is what the suite asks for:
| Function | Algorithm | Standard | Required parameter |
|---|---|---|---|
| Key establishment | ML-KEM (formerly CRYSTALS-Kyber) | FIPS 203 | ML-KEM-1024 |
| Digital signatures | ML-DSA (formerly CRYSTALS-Dilithium) | FIPS 204 | ML-DSA-87 |
| Symmetric encryption | AES | FIPS 197 | 256-bit keys |
| Hashing | SHA | FIPS 180-4 | SHA-384 or SHA-512 |
| Firmware and software signing (specific uses) | LMS and XMSS | NIST SP 800-208 | See the NSA FAQ for approved parameters |
NIST finalized FIPS 203, 204, and 205 in August 2024 (opens in new tab), and NIST describes them as ready for immediate use.
What Happens After January 2027?
The FAQ lays out a runway, and 2027 is the start of it, not the finish:
| Date | What NSA's FAQ says |
|---|---|
| January 1, 2027 | CNSA 2.0 algorithms required in all new NSS acquisitions. |
| During 2027 | NIAP stops accepting products for certification that have not started their quantum transition, and begins archiving certificates of products that do not conform. |
| December 31, 2030 | NSS equipment and services that cannot support CNSA 2.0 must be phased out unless granted an exception. |
| December 31, 2031 | CNSA 2.0 algorithms mandated for use to achieve full quantum resistance. |
Is There Any Flexibility on the Deadline?
A little, and it comes with conditions. NSA acknowledges that fully certified products may not exist in every category by the end of 2026. In those cases it may accept a purchase based on a vendor's contractual commitment to deliver a CNSA 2.0 update, paired with a realistic deployment plan that meets the 2030 deadline. That arrangement has to be tracked and reported. NSA also warns that some components, firmware update mechanisms in particular, likely cannot be changed after purchase, so buyers should confirm those are quantum-resistant on day one.
Is the Pressure Only Coming From NSA?
No. On October 27, 2026, NIST finalized its transition report, NIST IR 8547 (opens in new tab). NIST says Executive Order 14412 requires federal high-value assets and high-impact systems to migrate by 2030 for key establishment and 2031 for digital signatures, and that many organizations working with the federal government will be affected too.
Why Does Entropy Matter for Post-Quantum Cryptography?
Entropy is the unpredictability a system draws on to create a key. Every key, nonce, and seed starts there. A pseudo-random number generator (PRNG) stretches a small seed into a long stream of numbers that look random, and anyone who learns the seed can replay the stream. A true random number generator (TRNG) measures a physical process instead, so there is no seed to recover. We cover the difference in random vs. pseudorandom number generators.
Post-quantum algorithms change the math. They do not change the dependency. ML-KEM and ML-DSA key generation both begin with random seed values, and the standards expect those values to come from an approved random bit generator. A lattice-based key built from a weak seed is still a weak key. Quantum resistance protects you from one kind of attacker. It does nothing about an attacker who can guess your seed.
What Does Weak Entropy Look Like in the Real World?
The 2012 survey of TLS and SSH servers is still the clearest picture. The researchers found that 0.75% of TLS certificates shared keys because of insufficient entropy during key generation (opens in new tab), and that shared factors let them compute RSA private keys for 0.50% of TLS hosts. Most of the vulnerable hosts were headless or embedded devices, and the team traced part of the problem to a boot-time entropy hole in the Linux random number generator.
The pattern matters more than the percentages. Failures cluster where entropy is scarcest: first boot, small devices, and systems that start from identical images. Those conditions still exist in cloned virtual machines and containers launched by the thousand. For a recent example of weak key generation at scale, see how weak entropy led to the LuBian bitcoin theft. For the longer argument, read why true random number generation matters in the post-quantum era.
Where Does Entropy Fit in a CNSA 2.0 Compliance Plan?
CNSA 2.0 names algorithms. Entropy lives in the layers underneath them, and each layer has its own NIST document. Here is the stack from the bottom up, with the question worth asking a vendor at each level:
| Layer | What it does | Governing document | Question to ask |
|---|---|---|---|
| 1. Entropy source | Produces raw, unpredictable bits | NIST SP 800-90B (opens in new tab) | Is there 800-90B evidence for this source? |
| 2. Conditioning and health tests | Removes bias and flags a failing source | SP 800-90B | What happens when a health test fails? |
| 3. DRBG | Expands a seed into many random bits | SP 800-90A | Which DRBG, and how often does it reseed? |
| 4. RBG construction | Combines source and DRBG into a full generator | NIST SP 800-90C (opens in new tab) | Which RBG class does the module implement? |
| 5. Key generation | Creates ML-KEM-1024, ML-DSA-87, and AES-256 keys | FIPS 203, 204, 197 | Which generator feeds key creation? |
| 6. Validated product | Holds all of it in a certified boundary | CMVP and NIAP | Does the validation cover the entropy source? |
The NIST names are worth decoding. A DRBG is a deterministic random bit generator, the algorithmic part. An RBG is the complete generator. SP 800-90B covers the entropy source feeding it, and SP 800-90C, finalized in September 2025, specifies how the pieces are assembled into four RBG classes: RBG1, RBG2, RBG3, and RBGC.
What Does NSA Actually Say About Random Number Generators?
The CNSA 2.0 FAQ answers this directly. It states that, "regardless of quantum or traditional, all commercial RNGs used in NSS must generally at least satisfy NIST SP 800-90 A, B, and C." It also says any properly certified or approved RNG should be acceptable within the constraints of that approval.
There is a second piece on the NIST side. Since November 7, 2020, the Cryptographic Module Validation Program (CMVP) requires every FIPS 140-2 and FIPS 140-3 submission to include documentation of conformance to SP 800-90B (opens in new tab) where it applies. NIST also runs Entropy Source Validation (ESV), which issues certificates for sources that pass.
Does CNSA 2.0 Require a True Random Number Generator?
No. The suite does not name a TRNG, and the FAQ says CNSS policy does not favor quantum RNGs over other types. What it requires is evidence: a generator that meets the SP 800-90 series and sits inside a validated product. A physical entropy source is one of the cleanest ways to produce that evidence, because the source can be measured and tested directly. But the burden of proof is on the documentation, not the marketing.
Why Use a Physical Entropy Source for CNSA 2.0 Workloads?
Software runs on deterministic hardware, so it can only gather randomness from outside events: timing jitter, disk activity, network arrivals. On a server with a busy kernel that works. On a freshly booted appliance, a cloned virtual machine, or a small embedded device, those events are scarce, and that is exactly where the 2012 survey found trouble. A physical source avoids the problem by measuring a process that does not depend on the computer's own state.
SP 800-90B is built around that idea. It asks for a documented noise source, health tests that catch failure, a min-entropy estimate, and a conditioning step. A source whose physics you can describe and watch gives an evaluator something concrete to test, which is why hardware random number generators are the usual answer for high-assurance key generation.
How Does Real Random Produce Entropy?
Dice tumble in viscous fluid inside a sealed chamber. Optical readers capture each roll as raw pixel data, and the data is hashed into 256-bit or 512-bit values, about one per second, while the DieHarder test suite checks the output live. Values are single-use and deleted when drawn. You can see the process running, which we explain further in why the digital world needs physical entropy.
The measured results are public: 0.999996 bits per byte in Fourmilab ENT testing (opens in new tab), against a theoretical maximum of 1.0, with independent evaluation by Rochester Institute of Technology. NIST SP 800-90B entropy source validation is in process at atsec. That is a status, not a certificate, and we describe it that way on purpose.
Where Does It Connect to Your Stack?
An entropy source sits under everything that creates keys: the operating system pool, crypto libraries like OpenSSL and wolfSSL, HSMs, key management systems, and internal certificate authorities. It can run as an on-premises appliance behind your firewall or as Entropy-as-a-Service over an API. If the feed is ever unreachable, systems keep running on their native generator. The full walkthrough is on our entropy in cryptography guide.
One caution so nobody over-reads this. An entropy source is one component. It does not make a product CNSA 2.0 compliant on its own. The product still needs the right algorithms and the right validation around them.
What Should Teams Do Before January 2027?
This checklist fits in the time left before January 1, 2027, and none of it requires replacing your crypto libraries.
- Confirm scope. Do you own, operate, or sell into National Security Systems? If yes, the January 1 date applies to your next acquisition or bid.
- Inventory where keys are born. List every server, VM, container, HSM, key management system, certificate authority, and device fleet that generates keys, plus any application code calling a random function directly.
- Identify what feeds each one. Flag anything seeded at first boot, cloned from an image, or running on embedded hardware. Those are your highest-risk spots.
- Ask vendors for evidence, not adjectives. Request their CNSA 2.0 roadmap, their CMVP and NIAP status, and SP 800-90B documentation for the entropy source.
- Read the contract for update commitments. If a product will gain CNSA 2.0 support later, get the delivery date in writing and confirm firmware signing is quantum-resistant at purchase.
- Test and monitor. Run NIST and DieHarder-style statistical tests on generator output, and alert on health-test failures the way you alert on uptime.
- Plan for hardware refresh. NSA says equipment transitions should finish by December 31, 2030, and some products will need replacing, not patching.
Not sure where your stack stands? The Entropy Scorecard is a quick way to find the thin spots.
What Is the Bigger Picture?
CNSA 2.0 will be remembered as an algorithm story, but the algorithms are the visible half. The other half is the supply of unpredictable bits that every one of those algorithms consumes, and that supply has to be documented, tested, and validated like everything else on the list. Teams that treat entropy as a line item now will have an easier 2027 than teams that discover it during an audit. Post-quantum security is the math plus the randomness underneath it, and the second part deserves the same attention as the first. For the foundations, start with what post-quantum cryptography is.
Summary
- The date: Starting January 1, 2027, NSA's CNSA 2.0 algorithms are required in all new acquisitions for National Security Systems (NSS), according to NSA's CNSA 2.0 FAQ (opens in new tab).
- The suite: ML-KEM-1024 for key establishment, ML-DSA-87 for signatures, AES-256 for encryption, and SHA-384 or SHA-512 for hashing.
- The gap: None of those algorithms creates its own randomness. Every key still starts as random bits, and CNSA 2.0 does not tell you where they come from.
- The rule that does: NSA says commercial random number generators used in NSS must generally satisfy NIST SP 800-90 A, B, and C. It also says CNSS policy does not favor quantum RNGs over other types.
- Your move before January: Find every place your systems generate keys, identify what feeds each one, and ask vendors for SP 800-90B evidence, not just an algorithm list.

