ML-KEM v1.1.0
This release introduces a substantial update to the userspace implementation while preserving API compatibility.
Highlights. Userspace implementation improvements
- Reduced decapsulation overhead by eliminating repeated public-key unpacking.
- Moved critical modular arithmetic operations to inline implementations.
- Removed redundant memory zeroization operations.
- Additional SHA3 / SHAKE and Keccak optimizations.
Benchmarking and evaluation
A dedicated benchmark suite has been added, including:
- throughput measurements;
- stack usage analysis;
- memory usage measurements;
- comparisons against PQClean ML-KEM implementations.
Notes
- The Linux kernel implementation remains experimental and currently reflects an earlier stage of development. The userspace implementation is the primary target of ongoing development and validation.
- This release focuses on improving performance, reducing runtime overhead, and expanding validation while preserving the existing public API.