Cryptographic algorithm benchmarking in mobile devices

The aim of this thesis was to determine the execution times of different cryptographic algorithms in one of the hardware used in Asha families and compare the execution times between HW accelerated, OpenSSL and a company proprietary cryptographic library. Moreover, the motivation was to find out if the HW accelerated cryptographic function should be used when available due to execution time. Furthermore the aim was to find out if the transition to use OpenSSL is to be preferred in terms of the execution times over the company proprietary cryptographic library. In order to give wider perspective of the cryptography the thesis introduces a brief history of the cryptography. The following cryptographic functions are discussed in the thesis: hash functions, message authentication, symmetric and asymmetric cryptographic algorithms. The cryptographic algorithms are also examined from the embedded systems point of view. In addition of that the security of the embedded systems is discussed briefly in order to introduce the context where cryptographic functions are used. The other aim and the author’s personal motivation was to learn and deepen the knowledge of cryptographic algorithms, their usage and the issues to take into consideration when working with embedded devices. The research methods were a literature review and empirical research. The results supports hypothesis that it’s not always beneficial to use HW accelerated algorithms. The results surprised partially because the expectations were that HW accelerated execution would be much faster. Further investigations should be made in order to find the root cause for the results. Most of the algorithms were faster with OpenSSL in comparison to the company proprietary library and some of the HW accelerated algorithms. The performance is better in most of the crypto operations with OpenSSL than with the company proprietary crypto library. The thesis contributes the fact that no one should assume that it is obvious that the HW accelerated crypto operations are always faster. In many cases it is not significant in terms of performance, but in those cases where the performance is the critical issue the execution times should be measured rather than assumed