Benchmarks are hard. Cryptographic benchmarks are harder. 📊⏱️
New post-quantum algorithms like ML-DSA can be incredibly difficult to benchmark, because their runtimes can vary depending on random values, meaning lots of time-consuming tests are needed to estimate performance.
After sharing how we achieved ~3x performance on core RSA operations, we’re now giving a tour of how our cryptographic library can support a root of trust in everything from IoT to a BMC.
Building on our post from last week, we’re sharing the technical details behind our 3.43x RSA speedup from implementing Chinese Remainder Theorem (CRT) modular exponentiation in our embedded cryptography library.