HW-BLAST: High-Throughput Side-Channel TVLA and CPA Datasets

This repository contains side-channel power analysis datasets for evaluating secure cryptographic modules using Test Vector Leakage Assessment (TVLA) and Correlation Power Analysis (CPA). The data compares two distinct acquisition methodologies: a traditional manual capture over UART, and a high-throughput hardware-accelerated capture utilizing the HW-BLAST architecture. HW-BLAST accelerates the trace capture process by tightly coupling the target execution with an external oscilloscope. File Naming Convention The datasets are categorized by the acquisition mode used during physical capture: manual_*.h5: Baseline power traces acquired using the standard, unaccelerated trigger-and-capture methodology. blast_*.h5: The final, post-processed power traces after extracting and aligning the individual executions from the raw burst trace using the blast method. Dataset Contents & Trace Counts All datasets are provided in natively compressed HDF5 (.h5) format. File Name Experiment Trace Count Description *_cpa_aes.h5 CPA (AES-128) 30,000 Unprotected AES traces for evaluating Correlation Power Analysis attack convergence. *_tvla_aes.h5 TVLA (AES-128) 10,000 Fixed vs. Random traces for AES-128 encryption evaluation. *_tvla_eddsa.h5 TVLA (EdDSA) 10,000 Fixed vs. Random traces for EdDSA-25519 signature evaluation. *_tvla_sha512.h5 TVLA (SHA-512) 10,000 Fixed vs. Random traces for SHA-512 hashing evaluation. *_tvla_smaesh.h5 Masked TVLA (SMAesH) 2,500,000 Massive dataset of HPC2-masked AES traces. Used to evaluate higher-order statistical integrity.