Accelerating MD5 with CUDA on GPU

In order to improve the computing solution of cryptography, the possibility of using alternative computing solution based on GPU is studied in this paper. The use of a graphic processing unit in non-graphical calculations is tried to run the MD5 algorithm. We tried to use the graphic processing unit as a cryptographic coprocessor in order to obtain more computing power and better runtime for MD5. In this paper we present an implementation of MD5 on NVIDIA GPU using CUDA. The results of our tests show that the CUDA implementation can offer speedups of almost 50 times in comparison to the CPU. The tests are conducted in two directions: running the tests on small amount of data that is located in memory and a big amount of data that are stored in files on hard drives and as news, the access time for the information on the hard disk is added to the encrypting time.