365 Architect

06 — Security Levels

Two Parameter Sets

FN-DSA is simpler than ML-DSA — only two security levels instead of three:

FN-DSA-512 FN-DSA-1024
NIST Security Level Level 1 (~AES-128) Level 5 (~AES-256)
Ring dimension (n) 512 1,024
Modulus (q) 12,289 12,289
Public key 897 B 1,793 B
Private key 1,281 B 2,305 B
Signature 666 B 1,280 B
Sign time ~0.5 ms ~1.2 ms
Verify time ~0.08 ms ~0.15 ms

How the Parameters Work

Ring Dimension (n)

The dimension of the polynomial ring R_q = Z_q[X]/(X^n + 1):

  • n = 512: Polynomials have 512 coefficients
  • n = 1,024: Polynomials have 1,024 coefficients

Higher n means:

  • More security: Larger lattice = harder to find short vectors
  • Larger keys: 2× more coefficients = 2× larger public key
  • Larger signatures: 2× more coefficients = ~2× larger signature
  • Slower operations: 2× more FFT points = ~2.2× slower (log factor)

Modulus (q = 12,289)

Both parameter sets use the same modulus because:

  • q must support NTT up to n = 1,024 (actually up to 4,096)
  • 12,289 = 2^12 × 3 + 1 = 4,096 × 3 + 1
  • 4,096 = 212, so q ≡ 1 (mod 212), supporting NTT up to 4,096 points
  • Fits in 16-bit integers (q < 2^16 = 65,536)

Gaussian Width (σ)

The standard deviation of the discrete Gaussian distribution:

  • Both parameter sets use σ ≈ 1.17 × √q (normalised)
  • This is tight — signatures are as small as possible while maintaining security

Security Analysis

Classical Attacks

Attack Cost against FN-DSA-512 Cost against FN-DSA-1024
Lattice reduction (BKZ) ~2128 ~2256
Meet-in-the-middle ~2256 ~2512
Primal attack ~2128 ~2256
Dual attack ~2130 ~2258

Quantum Attacks

Attack Cost against FN-DSA-512 Cost against FN-DSA-1024
Grover on hash ~2128 ~2256
Quantum BKZ ~264 (estimated) ~2128 (estimated)
Quantum sieve ~290 (estimated) ~2180 (estimated)

Important caveat: Quantum lattice attacks are not well-understood. The estimates above are conservative; actual security may be higher.

How to Choose

Default: FN-DSA-512

Use this for:

  • Standard security requirements (equivalent to AES-128)
  • Bandwidth-constrained environments (666-byte signatures)
  • Performance-critical applications (faster signing)

When is AES-128 enough?

  • Short-term signatures (session tokens, ephemeral certificates)
  • Environments where keys are rotated frequently
  • Systems where the cost of Level 5 is prohibitive

Use FN-DSA-1024 When

  • Long-term signatures (legal documents, archive notarisation)
  • High-assurance environments (root certificates, trust anchors)
  • Regulatory requirements mandate highest security level
  • Future-proofing against advances in quantum algorithms

When is AES-256 necessary?

  • National security systems
  • Long-term legal evidence
  • Financial systems with multi-decade lifetimes
  • Scenarios where algorithm replacement is physically impossible

Comparison with Other Schemes at Same Levels

Level 1 (AES-128 equivalent)

Algorithm Public key Signature Sign Verify
FN-DSA-512 897 B 666 B ~0.5 ms ~0.08 ms
ML-DSA-44 1,312 B 2,420 B ~0.3 ms ~0.08 ms
SLH-DSA-SHA2-128s 32 B 7,856 B ~8 ms ~0.8 ms
ECDSA P-256 32 B 64 B ~0.05 ms ~0.1 ms

FN-DSA-512 has the smallest signature among PQC schemes at Level 1.

Level 5 (AES-256 equivalent)

Algorithm Public key Signature Sign Verify
FN-DSA-1024 1,793 B 1,280 B ~1.2 ms ~0.15 ms
ML-DSA-87 2,592 B 4,595 B ~0.8 ms ~0.18 ms
SLH-DSA-SHAKE-256s 64 B 29,792 B ~40 ms ~2 ms
ECDSA P-521 66 B 132 B ~0.2 ms ~0.3 ms

FN-DSA-1024 has the smallest signature among PQC schemes at Level 5.

Resources

  • NIST FIPS 206, Section 4: Parameter Sets
  • Security analysis: falcon-sign.info/security.html
  • Albrecht et al., "Estimate all the {LWE, NTRU} schemes!" (2018), SCN
Share on LinkedIn