International Conference on Computer Communications and Networks (ICCCN)
On the Effect of Post-Processing on QRNG-Based Key Generation for Resource-Constrained IoT
Oluwole Adetifa, Guobin Xu, Jianzhou Mao, Eric Sakk, Shuangbao Wang
Abstract
High-quality entropy is critical for secure key generation in resource-constrained Internet of Things (IoT) systems. While hardware Quantum Random Number Generators (QRNGs) are increasingly adopted, many systems still apply post-processing by default without first verifying whether these steps improve or degrade the quality of an already strong quantum source. This paper presents an empirical study comparing raw and post-processed QRNG output for cryptographic key generation, using an ID Quantique PCIe-240M accessed directly via the kernel driver in raw Sample Mode, bypassing the vendor library. Across 30 independent datasets, including a 24-cycle hourly study, we perform the evaluations with NIST SP 800-22 testing, Shannon entropy measurement, byte frequency analysis, bias and correlation checks, XOR shifted-copy autocorrelation testing, and Kolmogorov-Smirnov testing. Two conditioning methods, Von Neumann corrector and XOR whitening, are implemented as VHDL modules and validated through FPGA simulation in GHDL to enable controlled conditioning experiments. Our evaluation results show that post-processing degraded the QRNG source. XOR whitening reduced entropy and introduced measurable bias, while the Von Neumann corrector consistently failed NIST STS evaluation. QRNG-derived keys were further integrated into an authenticated encryption workflow using AES-256-CBC with HMAC-SHA256 to demonstrate practical applicability in IoT.