For SPECK128, optimal differential weights were known up to 11 rounds; for 12 rounds the best published characteristic has weight 66 (Kurahara, Sakamoto and Isobe, IEICE Trans. 2025), but whether a lighter one exists was open. We split the question into 36 ranges and showed that none has a characteristic of weight 65 or less, so 66 is optimal.
How it was shown
The search for a characteristic of weight 65 or less was split into 36 ranges, j15 to j50, and each range was shown to have no solution. The result is reported on the research index with the per-range outcome; the search inputs and solver logs are not yet published, and this page will link to them when they are. An independent re-check with a second encoding and a second solver is running.
Cite as
Nobuo Otoi (ba-cybersecurity.com). SPECK128, 12 rounds: no characteristic has weight 65 or less (weight 66 is optimal). 2026. https://ba-cybersecurity.com/research/speck128-12-rounds/.
@misc{otoi2026speck128,
author = {Otoi, Nobuo},
title = {SPECK128, 12 rounds: no characteristic has weight 65 or less (weight 66 is optimal)},
year = {2026},
howpublished = {\url{https://ba-cybersecurity.com/research/speck128-12-rounds/}}
}
Result listed on the research index 6 Nov 2026 · This page: updated 7 Nov 2026 · Permanent address: https://ba-cybersecurity.com/research/speck128-12-rounds/