Parallelizable Message Preprocessing for Merkle-Damgård Hash Functions

Since well-known hash functions sequentially process a message, the time for computing a digest strongly depends on the performance of a single processor. Even if multi-core processors are available, it is difficult to reduce the time. This paper focuses on designing a message preprocessing that can fully utilize multi-core processors. Since our message preprocessing can be efficiently computed by massively parallel processing, it contributes to reduce the time for computing a digest. The experiments using graphics processing units show that our message preprocessing improves the throughput for computing a digest.