[Nguyen* 6(7): July, 2019] ISSN 2349-4506 Impact Factor: 3.799 Global Journal of Engineering Science and Research Management http: // www.gjesrm.com © Global Journal of Engineering Science and Research Management [18] RECOVERING MS OFFICE PASSWORD ON PARALLEL COMPUTING SYSTEMS WITH MULTI-CORE PROCESSORS Nguyen Minh Quy*, Nguyen Dinh Han, Do Anh Tuan, Nguyen Van Hau, Pham Quoc Hung, Nguyen Xuan Truong, Nguyen Van Quyet, Ho Khanh Lam * Information Technology Faculty, Hung Yen University of Technology and Education, Vietnam DOI: 10.5281/zenodo.3345111 KEYWORDS: Parallel processing, Password recovery, Performance analysis, Multi-cores programming. ABSTRACT Recovering MS Office password using the Brute-Force method with multi-core CPUs and GPU support has been studied in several works, and obtained some success. In these works, the combination of multi-core CPUs and computer clusters in the computing process is essential in order to increase the processing speed. In this paper, we propose a method for recovering Microsoft Office passwords by an exhaustive method based on a parallel computing system with multiple compute nodes and also take advantage of multi-core capabilities in each node. Moreover, this paper presents an assessment of the impact of communication delays on the performance of a parallel computing system. INTRODUCTION Nowadays, using parallel computing systems to solve the problems with high complexity and substantial computation cost have been attracted by many studies. These systems often use computational clusters with hundreds or even thousands of compute nodes for processing data, and each node has multiple CPU and GPU cores) involved in the computation process [1,2]. Analyzing the performance of parallel computing systems is a complex problem. Gropp et al. [3] divided the execution time of a parallel computing system into three parts: calculation time, communication time, and idle time while handling large datasets. Thomas Rauber and Gudula Rünger indicated that the advances of multi-core processors and cache technology could help to build parallel computing systems more efficiently [1]. Several approaches for analyzing the performance of parallel algorithms have been presented in [5] and [6]. Xian-He Sun et al. [7] proposed an approach based on the study by Amdahl in [5] for analyzing the performance analysis of multi-core systems. In this paper, we proposed a method of analyzing performance for parallel computing systems by improving the approach developed by Xian-He Sun et al. in [7]. We conduct experiments on a real multi-core system with two compute nodes for solving the password recovering problem. Recovering MS Office password is a particular problem which can be applied a parallel computing mechanism to improve the performance of a password recovery system. There have been several studies focused on solving this problem on multiple CPU cores and GPU cores [8, 9, and 10]. In this paper, we proposed an approach for recovering password of MS Office 2010 by combining multi-core CPUs in a compute node and multiple compute nodes in a cluster with unlimited scalability to create a two-level parallel computing system. The system does not require to compute nodes to have the same hardware settings. It can be a network of personal computers and servers which are connected to each other via a Wifi or LAN/WAN network. Besides, we also discuss the problem of communication delay, which affects the performance of the system. We found that the acceleration level of a parallel computing system is not always completely linear with the number of nodes. The rest of the paper is organized as follows. Section 2 presents our approach for analyzing the performance of a parallel computing system. Section 3 presents the recovering MS Office password problem, a design and implementation of a parallel computing system for solving password recovering problem. In Section 4, we evaluate our approach with the analysis of the experimental results and discussions. Finally, we conclude the paper in Section 5.