Analisis dan Implementasi Algoritme Advanced Encryption Standard (AES) Secara Paralel

Abstract

AES is an algorithm for symmetric key cryptography. AES is expected to replace Data Encryption Standard (DES) as a recognized standard for various applications. The complexity of AES encryption and decryption process is O(n), it has fast computation in both encryption and decryption but its execution time will increase as the input data increase. In this research the AES algorithm would be parallelized by dividing the data to each processor (domain decomposition) and by dividing the constituent computing of AES to each processor (functional decomposition), and their performance would be analyzed. Parallel implementation of AES algorithm uses Message Passing Interface (MPI). The objectives of this research are to measure and analyze the performance of parallel AES algorithm using performance metrics. The best result of the parallel AES algorithm in this research is in the decryption process using domain decomposition method. A file with 118.525 MB size has 127.227 seconds decryption execution time using sequential AES algorithm, whereas using 16 processors, its decryption execution time is 34.896 seconds. The speedup is 3.646 and the efficiency is 0.228. This result is not good enough because the speedup is not equal to the number of processors used and the efficiency is below one. This is because the AES decryption in this research has a fast computational process, yet it has a big overhead for data communication. The worst result of the parallel AES algorithm is in the decryption process using functional decomposition method. Using the same file and three processors, the decryption execution time is 168.089 seconds. The speedup is 0.755 and the efficiency is 0.252. This is due to the fact that computing process decomposition is not equally distributed among the processes and the use of MPI blocking communication routine which cause a big overhead for data communication. Both parallel algorithms used in this research are not cost-optimal