<![CDATA[Cryptographic hash functions form the preprocessing layer of Ethereum fraud detection pipelines, yet prior research evaluated hash performance and fraud detection as separate concerns, leaving practitioners without evidence-based guidance for selecting an algorithm that balances speed, energy efficiency, and cryptographic security. This study benchmarks SHA-256, SHA3-256, BLAKE2s-256, and BLAKE3 on a stratified sample of 6,396 fraud-labeled transactions from a public 9,841-record Ethereum dataset, measuring execution time, throughput, energy consumption, and avalanche effect. Its novelty is, to the best of our knowledge, the first inferential validation, on real fraud-labeled data, that the four algorithms are cryptographically equivalent in diffusion strength—combining non-parametric testing (Shapiro-Wilk, Mann-Whitney U, Kruskal-Wallis) with formal two one-sided equivalence tests (TOST) and replicating the result on an independent 454,289-transaction dataset—establishing that selection can be decided on efficiency grounds alone without sacrificing security. BLAKE2s-256 achieved the lowest execution time (0.65 µs), highest throughput (≈1.54 million ops/s), and lowest energy (0.029 mJ/op), while all algorithms converged near the 50% avalanche ideal with no significant security difference (H = 3.03, p = 0.387). BLAKE2s-256 attained the highest composite score (99.98/100) and, in an end-to-end pipeline test, improved batch screening throughput by up to 11.5%, confirming it as optimal for off-chain Ethereum fraud-detection preprocessing on processors without SHA hardware acceleration.]]>
