Source PAC Coding for Low-latency Secret Key Generation in Short Blocklength Regime
Lulu Song, Di Zhang, Tingting Zhang
TL;DR
This work tackles low-latency secret-key generation from limited source observations in short blocklength regimes by introducing a multilevel source polarization-adjusted convolutional (PAC) coding framework. The approach combines a rate-one convolutional outer code with a multilevel polarization scheme, and employs a Gaussian-approximation-based code-construction algorithm to balance polarization gains with ML decoding performance. A multistage SCL decoding strategy at the receiver and dynamic frozen bits enable effective reconciliation under short $N$, achieving higher key rates under reliability constraints than conventional and multilevel source polar codes. Simulation results demonstrate significant gains in $R_K$ at short blocklengths and competitive performance relative to finite-blocklength bounds, highlighting potential for low-latency key generation in 6G IoT scenarios.
Abstract
Source polar coding is a potential solution for short blocklength-based low-latency key generation with limited sources, which is a critical aspect of six generation (6G) Internet of things. However, existing source coding schemes still suffer from significant degradation in key generation rate and reconciliation reliability in short blocklength regime. To address this issue, we introduce a multilevel source polarization-adjusted convolutional (PAC) coding framework. Furthermore, we propose a novel code construction algorithm that jointly leverages polarization effects and the maximum likelihood (ML) decoding error coefficient. Simulations demonstrate that the multilevel source PAC scheme with the proposed code construction achieves superior key generation rate under key disagreement constraints compared to conventional and multilevel source polar coding methods even in short blocklength regimes.