Jorge Quitério Search
IT & Engineering

Planning the migration to post-quantum cryptography

The post-quantum standards are published. The main difficulty is identifying where cryptography is used and how quickly it can be replaced. Mosca's inequality, expressed as a probability, indicates when the migration should start.

A large quantum computer capable of breaking today's public-key cryptography does not exist yet. Governments are nevertheless already setting migration deadlines.

The reason is that part of the risk materialises now. An adversary can record encrypted traffic now and decrypt it once a capable machine exists: harvest now, decrypt later. Any secret that must stay secret for longer than the time left is already exposed.

Personal data shows why. A medical record or a mortgage agreement sent over the internet today must remain confidential for decades. If the encrypted traffic is recorded now, its confidentiality ends when a capable quantum computer appears, not when the patient or the borrower no longer cares about it.

In August 2024 NIST published its first three post-quantum standards, so choosing which algorithms is largely settled. The open questions are where cryptography is used in a system and how long it will take to replace. This article addresses when to start and what to do first.

Several standards and papers address this problem.

The threat. Shor (1994) showed that a quantum computer could factor integers and compute discrete logarithms in polynomial time, breaking RSA and elliptic-curve cryptography. Grover (1996) showed a quadratic speed-up for unstructured search, which weakens symmetric keys but does not break them.

When to act. Mosca (2018) framed the timing question as a simple inequality between data shelf life, migration time and the arrival of a cryptographically relevant quantum computer.

The standards. NIST published FIPS 203 (ML-KEM), FIPS 204 (ML-DSA) and FIPS 205 (SLH-DSA) in August 2024, and in March 2025 selected HQC as an additional key-establishment algorithm based on different mathematics.

The timelines. In Europe, a coordinated roadmap from the NIS Cooperation Group (June 2025) asks member states to start the transition by the end of 2026, protect high-risk use cases by the end of 2030 and migrate as much as feasible by 2035. The NSA's CNSA 2.0 sets transition dates for US national security systems.

Mosca's inequality

If xx is how long the data must remain confidential, yy how long the migration will take, and zz the time until a capable quantum computer exists, then

x+y>z x + y > z

means the migration is already late. Nobody knows zz. However, xx and yy can be estimated by the organisation, and yy is usually measured in years.

The standards

Standard Algorithm Purpose
FIPS 203 ML-KEM (from CRYSTALS-Kyber) Key establishment
FIPS 204 ML-DSA (from CRYSTALS-Dilithium) Digital signatures
FIPS 205 SLH-DSA (from SPHINCS+) Hash-based digital signatures

Symmetric cryptography is in a much better position: Grover's search roughly halves the effective key length, so AES-256 keeps about 128 bits of security, which remains adequate.

Cryptographic inventory

Cryptography that has not been identified cannot be migrated. The first deliverable is an inventory of where cryptography is used (protocols, libraries, certificates, keys, hardware security modules, embedded devices, third-party services), each with its algorithm, key size, owner and the lifetime of the data it protects. A cryptographic bill of materials (CBOM) formalises this; CycloneDX supports it since version 1.6. A spreadsheet is an adequate start, provided it is maintained.

Hybrid key exchange

Key exchange is what harvest-now-decrypt-later targets, so it should be addressed first. Hybrid schemes combine a classical algorithm with a post-quantum one, so security holds as long as either does. Modern browsers and TLS libraries support the hybrid group X25519MLKEM768, and enabling it at reverse proxies and load balancers protects traffic without touching applications. Recent OpenSSH releases offer hybrid key exchange too. Support moves fast, so current versions should be checked before planning.

Cryptographic agility

The broader requirement is crypto-agility: algorithm choices in configuration rather than code, a managed key lifecycle, and hardware security modules whose firmware can be upgraded. Signatures and certificates usually come last because whole ecosystems must move together. Systems with long lifetimes, such as industrial equipment or satellites, need agility designed in from the start, because they will outlive several generations of algorithms.

The figure below expresses the inequality as a probability. Because zz is uncertain, it makes more sense to ask how likely x+y>zx + y > z is than whether it holds. Long-lived data and slow migrations make that probability high even under optimistic assumptions about zz.

The figure treats zz as an adjustable belief: a median guess and a degree of uncertainty. The shaded area is the probability that a quantum computer arrives before data encrypted today stops being sensitive.

  1. Medical or state secrets. Set x=25x = 25 years and y=5y = 5. Even with a median guess of 30 years, the exposure is about 50%.
  2. Short-lived data. Set x=1x = 1 (a session token, say) and y=5y = 5. The risk falls sharply: this is why key exchange for long-lived data comes first, and short-lived authentication can wait a little.
  3. Faster migration. Each year removed from yy moves the line to the left. That is the case for starting the inventory now: it is the step that makes yy shorter.
  4. Greater uncertainty. Increase the uncertainty about zz. The probability of an early arrival grows, even with the same median. Uncertainty is therefore itself a reason to act early.

Calculation. With zz modelled as lognormal with median mm and log-scale spread ss, the exposure is P(z<x+y)=Φ((ln⁡(x+y)−ln⁡m)/s)P(z < x + y) = \Phi\big((\ln(x+y) - \ln m)/s\big), where Φ\Phi is the standard normal distribution function. The lognormal is my modelling choice and not a forecast: it keeps zz positive and lets experts express "probably in about this many years, but maybe much later".

Post-quantum migration is mainly an inventory and engineering project, with a smaller cryptographic component. It should start with a cryptographic bill of materials, move key exchange to hybrid schemes where this is inexpensive, and build in agility elsewhere. Mosca's inequality explains why the timeline has already started, and its probabilistic form gives the risk owner a quantity on which to base the decision.

References

  1. P. W. Shor, "Algorithms for quantum computation: discrete logarithms and factoring", Proceedings of the 35th Annual Symposium on Foundations of Computer Science, 124-134, 1994.
  2. L. K. Grover, "A fast quantum mechanical algorithm for database search", Proceedings of the 28th Annual ACM Symposium on Theory of Computing, 212-219, 1996.
  3. M. Mosca, "Cybersecurity in an era with quantum computers: will we be ready?", IEEE Security & Privacy, 16(5), 38-41, 2018.
  4. NIST, FIPS 203, Module-Lattice-Based Key-Encapsulation Mechanism Standard, August 2024.
  5. NIST, FIPS 204, Module-Lattice-Based Digital Signature Standard, August 2024.
  6. NIST, FIPS 205, Stateless Hash-Based Digital Signature Standard, August 2024.
  7. NIS Cooperation Group, A Coordinated Implementation Roadmap for the Transition to Post-Quantum Cryptography, June 2025.