Paper 2025/2293 Original: eprint.iacr.org/2025/2293 ↗

LAKE: Lattice-Code Accelerated Kyber Encapsulation

Abstract

The standardization of CRYSTALS-Kyber (ML-KEM) by NIST represents a milestone in post-quantum security, yet its substantial communication overhead remains a critical bottleneck for resource-constrained environments. This paper introduces <i>LAKE (Lattice-Code Accelerated Kyber Encapsulation)</i>, a novel cryptographic framework that symbiotically integrates coding theory into the Module-LWE structure. Unlike previous concatenation approaches, LAKE embeds density-optimized Construction-A lattices derived from Polar codes directly into the public matrix generation. This structural innovation yields a <i>15–25% reduction in ciphertext size</i> while simultaneously improving the Decryption Failure Rate (DFR) from \(2^{-139}\) to <i>\(2^{-156}\)</i>, leveraging innate coding gains to suppress noise. We provide a rigorous reduction of LAKE's IND-CCA2 security to the hardness of the Structured Module-LWE problem. Although LAKE introduces a modest 8–15% computational overhead, it optimizes the critical "Compute-for-Bandwidth" trade-off, exploiting the asymmetry between low-cost local processing and high-cost transmission. Consequently, LAKE significantly enhances deployment viability in high-latency, energy-sensitive domains such as Satellite Communications (SatCom), Narrowband-IoT (NB-IoT), and tactical edge networks, where transmission efficiency is the dominant performance metric.