On Software Implementation of High Performance GHASH Algorithms
MetadataShow full item record
There have been several modes of operations available for symmetric key block ciphers, among which Galois Counter Mode (GCM) of operation is a standard. GCM mode of operation provides confidentiality with the help of symmetric key block cipher operating in counter mode. The authentication component of GCM comprises of Galois hash (GHASH) computation which is a keyed hash function. The most important component of GHASH computation is carry-less multiplication of 128-bit operands which is followed by a modulo reduction. There have been a number of schemes proposed for efficient software implementation of carry-less multiplication to improve performance of GHASH by increasing the speed of multiplications. This thesis focuses on providing an efficient way of software implementation of high performance GHASH function as being proposed by Meloni et al., and also on the implementation of GHASH using a carry-less multiplication instruction provided by Intel on their Westmere architecture. The thesis work includes implementation of the high performance GHASH and its comparison to the older or standard implementation of GHASH function. It also includes comparison of the two implementations using Intel’s carry-less multiplication instruction. This is the first time that this kind of comparison is being done on software implementations of these algorithms. Our software implementations suggest that the new GHASH algorithm, which was originally proposed for the hardware implementations due to the required parallelization, can't take advantage of the Intel carry-less multiplication instruction PCLMULQDQ. On the other hand, when implementations are done without using the PCLMULQDQ instruction the new algorithm performs better, even if its inherent parallelization is not utilized. This suggest that the new algorithm will perform better on embedded systems that do not support PCLMULQDQ.
Cite this work
Iqbal Muhammad Umair (2012). On Software Implementation of High Performance GHASH Algorithms. UWSpace. http://hdl.handle.net/10012/7114