Optimizing Hash-Based Signatures in Java

Hash-based signature schemes are an extensively studied and well-understood choice for quantum-safe digital signatures. However, certain operations, most notably the key generation, can be comparably expensive. It is, therefore, essential to use well-optimized implementations.
This thesis aims to explore, implement, and evaluate optimization strategies for hashbased signature implementations in Java. These include the use of special hardware features like vector instructions and hardware acceleration for hash functions as well as the parallelization of the key generation. Overall, we are able to reduce the time required for an XMSS key generation with SHA-2 by up to 96.4% (on four CPU cores) compared to the unmodified BouncyCastle implementation. For SPHINCS+ with the Haraka hash function family, we achieve a reduction of up to 95.7% on only one CPU core. Furthermore, we investigate the use of two scheme variants WOTS-BR and WOTS+C proposed in the literature for verification-optimized signatures. We improve the existing theoretical analysis of both, provide a comparison and experimentally validate our improved theoretical analysis.