Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/4AC968CF-12A6-4C11-B9F9-3FDD59971B72 | ||
| Year | 2021 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Konstantinos Ampatzidis, "Hardware acceleration of adiantum cryptography algorithm on PYNQ", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2021 https://doi.org/10.26233/heallink.tuc.88981 | ||
| Appears in Collections | |||
| Relations with other Items | Is Referenced by the Item: |
As technology is closely interwoven with everyday reality, an exponentially increasing volume of information is exposed to potential data breaches. In that context, the field of cryptography offers the necessary confidentiality and accuracy to sensitive data handling. Existing options such as AES can often lead to significant performance expense, yet recent appearance of more lightweight alternatives, like Adiantum, resolves the dilemma of choosing between speed and security. The connection between hardware development and cryptography is inevitable as hardware offers high parallelism which in turn results in faster deployments with balanced power consumption. In this thesis we present the first attempt at accelerating the entire Adiantum algorithm for big plaintexts with FPGAs. This thesis comprises of three parts: profiling of the Adiantum algorithm in order to determine the most computationally intensive parts; implementation of the ChaCha12 core which accounts for some 86% - 96% of the total load; and, full implementation of the ChaCha12 core and the Adiantum algorithm on a PYNQ Z1 FPGA board. Despite technology-related limitations the results are most encouraging. Specifically, the ChaCha12 core is 10,731 times faster and 77,000 times more energy efficient than the Intel i5-3230M processor. If it were to run on a present-day system with an Intel i5-3230M processor having a tightly-coupled FPGA, the Adiantum algorithm, including I/O overhead would run at speeds approaching the theoretical limits posed by Amdahl's Law. However, because the processor on the PYNQ Z1 is 15 times slower than the Intel i5-3230M processor, the full-Adiantum-algorithm performance of the Intel i5-3230M CPU is 4x times faster than our Pynq-z1 system, but at a 2x higher energy cost.
Copyright © DIAS 2013
