PIPELINED DATA PARALLEL MODEL OF ADVANCED ENCRYPTION STANDARD ALGORITHM

Abstract

The Advanced Encryption Standard (AES) was officially adopted in 2002 as the new encryption standard algorithm. AES specifies a FIPS-approved cryptographic algorithm that can be used to protect electronic data. It is a symmetric block cipher that can encrypt and decrypt information. This paper develops a pipelined data parallel model of AES. The parallelism in the algorithm is two dimensional. The first dimension is AES enter-stage (pipelining) and the second dimension is data parallelism. Pipelining parallelism exploits the availability of several processes to execute different stages of different data blocks in parallel. The data parallelism exploits data independence among data blocks to implement data level parallelism. The parallel implementation of AES decreases the time needed for encryption and decryption processes. We use the ECB mode in encryption/decryption algorithm in our parallel implementation of AES to implement the parallelization at data level where data blocks are encrypted and decrypted in parallel. We also develop an MPI-based algorithm to be used with a cluster of workstations (COW). We validate the approach by simulating the model with various input parameters (input data file size, number of processes, communication/computation operation execution time, etc.) and measuring the corresponding performance. Performance metrics include speedup, communication to computation ratio and efficiency. Results show that performance obtained by the developed model is superior to parallel implementations of AES which include only data parallelism or pipelining.