Improvements in Brute Force Attacks

This blog post summarizes a 2025 research paper showing optimized GPU implementations (measured on RTX 4090) that substantially increase key-testing rates for KASUMI, SPECK, and TEA3, improving prior results (e.g., KASUMI >15× faster than prior work). The authors report per-GPU key rates and translate them into large-scale resource estimates (billions or millions of RTX 4090 GPUs to break respective protocols in a year) and note significant reductions in time/memory trade-off table creation when using modern GPUs; the work is important cryptanalytic research but remains impractical for real-world widescale exploitation at present.