Key length and PQC transition
Planning horizon

Key length

Data as of 8 October 2026. Every figure carries its own source and date.

The minimum key sizes each source publishes for protecting data until a given year. The rules that apply to you come first, and every figure links to the document it comes from. New to this? Start with why key length matters and which year to enter.

Rules that apply to you

No rules selected, so every source is listed as comparison.

Data encrypted today with RSA or elliptic-curve key exchange and needing secrecy until 2035 can be recorded now and decrypted later if a large quantum computer arrives first.

Any key is only as strong as its handling, because keys are stolen or leaked far more often than they are broken; that exposure is what cryptoperiods limit.

Equivalent classical security strength

Public-key parameter size is not the same thing as security strength. This table puts RSA, finite-field Diffie–Hellman or DSA, elliptic curves, hashes and an ideal symmetric key on the same scale. For example, 3072-bit RSA and a 256-bit elliptic curve both map to about 128 security bits.

Security bits
≈ symmetric key
RSA modulus Finite-field DH / DSA
modulus / subgroup
Elliptic-curve key Collision-resistant hash output
80 legacy1,0241,024 / 160160–223160
1122,0482,048 / 224224–255224
1283,0723,072 / 256256–383256
1927,6807,680 / 384384–511384
25615,36015,360 / 512512+512

Classical equivalence from NIST SP 800-57 Pt 1 Rev 5, Table 2. The 80-bit row is legacy only. These are conservative strength categories, not exact attack-cost estimates, and they do not make classical public-key cryptography resistant to a cryptographically relevant quantum computer.

How long to use a key

Key size answers how hard a key is to break; a cryptoperiod answers how long that particular key should remain authorized for use. Cryptoperiods exist because keys are stolen, leaked through bugs in the software around them, or used as signing oracles far more often than they are broken, so the period bounds the damage from a compromise rather than from cryptanalysis. These are NIST’s suggested starting points for rotation, not certificate validity limits or universal mandates. Is the algorithm usually the weak point?

Key type Suggested cryptoperiod
Signing
Private signature key1–3 years
Public signature-verification keySeveral years; depends on key size
Authentication
Symmetric authentication keyOriginator use <2 years; recipient use <originator period + 3 years
Private authentication key1–2 years
Public authentication key1–2 years
Encryption, wrapping and derivation
Symmetric data-encryption keyOriginator use <2 years; recipient use <originator period + 3 years
Symmetric key-wrapping keyOriginator use <2 years; recipient use <originator period + 3 years
Symmetric random-bit-generator keySee NIST SP 800-90
Symmetric master or key-derivation keyAbout 1 year
Private key-transport key<2 years
Public key-transport key1–2 years
Key agreement
Symmetric key-agreement key1–2 years
Private static key-agreement key1–2 years
Public static key-agreement key1–2 years
Private ephemeral key-agreement keyOne key-agreement transaction
Public ephemeral key-agreement keyOne key-agreement transaction
Authorization
Symmetric authorization key<2 years
Private authorization key<2 years
Public authorization key<2 years

From NIST SP 800-57 Pt 1 Rev 5, Table 1. NIST says the application and operating environment may justify shorter or longer periods; extending them requires explicit risk consideration. “Originator use” is when the key may protect new data, while “recipient use” is how long it may process already-protected data. An originator-use period is not a certificate lifetime: a two-year signing period does not mean a two-year certificate, and the public keys that verify roots, timestamped signatures and signed software run on a different, much longer clock than the private key that signed them. See What is a cryptoperiod?

NIST12031 to open
Symmetric
128
Factoring modulus
3,072
Discrete log
3,072 / 256
Elliptic curve
256
Hash
256
ML-KEM (FIPS 203)
Signatures: ML-DSA, SLH-DSA
BSI42026 to 2032
Symmetric
128
Factoring modulus
Hybrid only
Discrete log
Hybrid only
Elliptic curve
Hybrid only
Hash
256
Hybrid with a classical scheme: ML-KEM-768 or -1024, FrodoKEM-976 or -1344, or Classic McEliece (BSI advised on 1 October 2026 against Classic McEliece in new developments; revision of TR-02102-1 planned for early 2027)
Signatures: ML-DSA-65 or -87, SLH-DSA (192 or 256), LMS or XMSS; classical signatures until end of 2035
ANSSI52031 to open
Symmetric
128
Factoring modulus
3,072
Discrete log
3,072 / 250
Elliptic curve
250
Hash
256
Hybrid with a classical scheme: ML-KEM-768 preferred (512 allowed), FrodoKEM-976 preferred; post-quantum recommended for use beyond 2030
Signatures: ML-DSA hybrid with a classical signature; SLH-DSA may be used alone
NSA CNSA 2.062025 to open
Symmetric
256
Factoring modulus
Transition only
Discrete log
Transition only
Elliptic curve
Transition only
Hash
384
ML-KEM-1024
Signatures: ML-DSA-87; LMS or XMSS for firmware
ASD02031 to open
Symmetric
256
Factoring modulus
Retired
Discrete log
Retired
Elliptic curve
Retired
Hash
384
ML-KEM-768 or -1024, preferably 1024; only ML-KEM-1024 approved beyond 2030
Signatures: ML-DSA-65 or -87, preferably 87; only ML-DSA-87 approved beyond 2030
Source Window Symmetric ? Factoring modulus ? DL key / group ? Elliptic curve ? Hash ? Post-quantum ?
NIST12031 to open1283,0723,072 / 256256256ML-KEM (FIPS 203)
Sig: ML-DSA, SLH-DSA
BSI42026 to 2032128Hybrid onlyHybrid onlyHybrid only256Hybrid with a classical scheme: ML-KEM-768 or -1024, FrodoKEM-976 or -1344, or Classic McEliece (BSI advised on 1 October 2026 against Classic McEliece in new developments; revision of TR-02102-1 planned for early 2027)
Sig: ML-DSA-65 or -87, SLH-DSA (192 or 256), LMS or XMSS; classical signatures until end of 2035
ANSSI52031 to open1283,0723,072 / 250250256Hybrid with a classical scheme: ML-KEM-768 preferred (512 allowed), FrodoKEM-976 preferred; post-quantum recommended for use beyond 2030
Sig: ML-DSA hybrid with a classical signature; SLH-DSA may be used alone
NSA CNSA 2.062025 to open256Transition onlyTransition onlyTransition only384ML-KEM-1024
Sig: ML-DSA-87; LMS or XMSS for firmware
ASD02031 to open256RetiredRetiredRetired384ML-KEM-768 or -1024, preferably 1024; only ML-KEM-1024 approved beyond 2030
Sig: ML-DSA-65 or -87, preferably 87; only ML-DSA-87 approved beyond 2030

All sizes in bits and all are minimums. Grey italic values fall outside the source's own window and show where it last stood. D draft, S superseded edition, 2nd not yet checked against the primary document. The old equation-based estimators are intentionally kept separate from current policy; read why.

References

  1. NIST SP 800-57 Pt 1 Rev 5, NIST, 2020-05.
  2. NIST SP 800-57 Pt 1 Rev 6 initial public draft, NIST, 2025-12-05, draft.
  3. NIST IR 8547 initial public draft, NIST, 2024-11-12, draft.
  4. BSI TR-02102-1, Cryptographic Mechanisms: Recommendations and Key Lengths, version 2026-01, BSI, 2026-01-23.
  5. ANSSI-PG-083, Règles et recommandations concernant le choix et le dimensionnement des mécanismes cryptographiques, v3.00, ANSSI, 2026-03-20.
  6. NSA CNSA 2.0 algorithms advisory, NSA, 2025-05-30.
  7. ASD, Planning for post-quantum cryptography, ASD, 2025-09-22.

How your rules compare

Each line is one source's minimum over the years it covers, with your rules in colour. The shaded band spans the least and most demanding sources and the dashed line is the median.

RetiringQuantum-vulnerablealgorithms retired1,5362,0482,5603,0723,58420202022202420262028203020322034203620382040NISTANSSIBSIASD2035
Range across sourcesMedianYour rules, in colourOther sourcesClassical retirement