Performance analysis of a scalable hardware FPGA Skein implementation.

摘要

Hashing functions are a key cryptographic primitive used in many everyday applications, such as authentication, ensuring data integrity, as well as digital signatures. The current hashing standard is defined by the National Institute of Standards and Technology (NIST) as the Secure Hash Standard (SHS), and includes SHA-1, SHA-224, SHA-256, SHA-384 and SHA-512. SHS's level of security is waning as technology and analysis techniques continue to develop over time. As a result, after the 2005 Cryptographic Hash Workshop, NIST called for the creation of a new cryptographic hash algorithm to replace SHS. The new candidate algorithms were submitted on October 31st, 2008, and of them fourteen have advanced to round two of the competition. The competition is expected to produce a final replacement for the SHS standard by 2012.;Multi-core processors, and parallel programming are the dominant force in computing, and some of the new hashing algorithms are attempting to take advantage of these resources by offering parallel tree-hashing variants to the algorithms. Tree-hashing allows multiple parts of the data on the same level of a tree to be operated on simultaneously, resulting in the potential to reduce the execution time complexity for hashing from O(n) to O(log n). Designs for tree-hashing require that the scalability and parallelism of the algorithms be researched on all platforms, including multi-core processors (CPUs), graphics processors (GPUs), as well as custom hardware (ASICs and FPGAs). Skein, the hashing function that this work has focused on, offers a tree-hashing mode with different options for the maximum tree height, and leaf node size, as well as the node fan-out.;This research focuses on creating and analyzing the performance of scalable hardware designs for Skein's tree hashing mode. Different ideas and approaches on how to modify sequential hashing cores, and create scalable control logic in order to provide for highspeed and low-area parallel hashing hardware are presented and analyzed. Equations were created to help understand the expected performance and potential bottlenecks of Skein in FPGAs. The equations are intended to assist the decision making process during the design phase, as well as potentially provide insight into design considerations for other tree hashing schemes in FPGAs. The results are also compared to current sequential designs of Skein, providing a complete analysis of the performance of Skein in an FPGA.