Skip to content

Back to the catalog

SLEEK: Sparse Quasi-Cyclic Codes with Closed-Form DFR Key Encapsulation
eprint.iacr.orgpaper

SLEEK: Sparse Quasi-Cyclic Codes with Closed-Form DFR Key Encapsulation

Introduces SLEEK, a QC-MDPC code-based KEM using a three-iteration parallel bit-flipping decoder and a closed-form decoding-failure-rate (DFR) estimate, enabling parameters with analytically bounded failure probability and standard IND-CCA transforms. SLEEK-LT targets IND-CCA for long-term keys; SLEEK-E targets IND-CPA for ephemeral keys using a new QC key structure with security equivalent to QC-MDPC and cheaper key generation. Parameter sets cover 80/128/192/256-bit security, and the optimized AVX2 implementation gives ciphertexts 4.9x smaller than HQC (47% slower) and 1.3x–2.7x faster than BIKE.

TL;DR SLEEK is a QC-MDPC post-quantum KEM with closed-form DFR bounds that beats BIKE on speed and HQC on ciphertext size in AVX2 benchmarks.
Open original ↗
Document

What it is

SLEEK is a QC-MDPC code-based KEM from Annechini, Barenghi, Gianvecchio and Pelosi (Politecnico di Milano). It is an October 2026 ePrint preprint, not peer reviewed. The only text available here is the abstract and metadata, so nothing below comes from the paper body.

Variants and what the summary leaves out

  • SLEEK-LT (IND-CCA, long-term keys): supports QC-MDPC codes at several rates, so you can trade public-key size against ciphertext size. Its conservative choices are aimed at active attackers.
  • SLEEK-E (IND-CPA, ephemeral keys, forward-secrecy settings): uses a new QC key structure. The authors claim security equivalent to original QC-MDPC under the same assumptions, with cheaper key generation.

Benchmarks (the authors' optimized AVX2 implementation)

ComparisonResult
SLEEK-LT vs HQCCiphertext 4.9x smaller, runtime +47%
SLEEK-LT vs BIKE1.3x to 2.7x faster
SLEEK-E vs HQC1.6x faster, bandwidth 2.1x lower
SLEEK-E vs Lean-BIKE5.3x faster, less impact from decoding failures on system reliability

The authors also say they optimized polynomial inversion and the parallel bit-flipping decoder, which they call the two dominant implementation costs. The 80-bit set is meant for constrained embedded platforms. The other sets target 128, 192 and 256 bits.

Caveats

  • The closed-form DFR estimate is what justifies using standard IND-CCA transforms. Its validity and tightness can't be judged from the abstract. Check the derivation before relying on the claimed failure bounds.
  • The SLEEK-E security equivalence is the authors' own claim, made "under the same assumptions" as QC-MDPC.
  • The comparisons use the authors' own AVX2 code. The abstract doesn't say which BIKE and HQC implementations were used as baselines.
  • HQC is called the NIST standard choice, while SLEEK is not standardized. It is a research candidate, not a drop-in replacement.

Organisations named here

  • QC-MDPC
  • SLEEK
  • SLEEK-E
  • SLEEK-LT

Technologies named here

  • AVX2
  • LeanBIKE
  • HQC
  • BIKE
  • NIST

Techniques named here

  • Parallel bit-flipping decoding with three iterations
  • Standard IND-CCA transformations enabled by bounded DFR
  • Optimized polynomial inversion

Also named here

  • Lean-BIKE

Related in other