Performance Evaluation of Twofish Algorithm on IMAN1 Supercomputer

Full text

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
1
Performance Evaluation of Twofish Algorithm on IMAN1
Supercomputer
Heba Harahsheh
Department of Computer Science, King Abdullah II
School for Information Technology
University of Jordan, Amman, Jordan
Mohammad Qatawneh
Department of Computer Science, King Abdullah II
School for Information Technology
University of Jordan, Amman, Jordan
ABSTRACT
Now-a-days data is needed to be exchanged through networks
in secure manner. Hence for secure communication required
cryptography algorithms. Among these algorithms is Twofish
cryptographic algorithm. In this paper the Sequential and
parallel implementation have been implemented in IMAN1
supercomputer using Message Passing Interface (MPI). The
parallel implementation has been evaluated in terms of
execution time, speedup, and efficiency. The simulation
results show that 256 key lengths in 4 processors get best
efficiency up to 37% while 192 and 128 key length achieves
up to 33% and 29% in order.
Keywords
Twofish Algorithm, Parallel, Supercomputer, MPI.
1. INTRODUCTION
Nowadays many systems need huge complex computations in
many fields such as industry, and all of these computations
use parallel computing in order to get more performance [1].
Parallel and distributed computing systems divide large
problems into smaller sub-problems and assign each of them
to different processors in a typically distributed system
running concurrently in parallel [2] [3] [4] [5] [6].
Transforming data through internet is critical. To secure data
transaction and communications, we need to use
cryptographic algorithms. Several cryptography algorithms
have proposed like Twofish, AES, DES, 3DES, RC2 [7] [8]
[9] [10]. Among these algorithms is Twofish cryptography
algorithm [11] [12]. Cryptosystems has two types Symmetric
Key Encryption (using same key for encryption and
decrypting) and Asymmetric Key Encryption (different keys
are used for encrypting and decrypting the information). The
security of encryption increasing depends on the secure key
and strength of cryptographic algorithm.
In this paper, the Performance Evaluation of Twofish
algorithm on Supercomputer IMAN1 is presented. IMAN1 is
Jordan's first and fastest High Performance Computing
resource, funded by JAEC and SESAME [1]. It is available
for use by academia and industry in Jordan and the region.
The evaluation is done in terms of the running time, speed and
parallel efficiency according to different data size and
different number of processors. The results were conducted
using IMAN1. It is available for use by academia and industry
in Jordan and the region and provides multiple resources and
clusters to run and test High Performance Computing (HPC)
codes [1].
This paper is organized as follows: Twofish algorithm and
related works are presented in Section 2. In section 3,
presents the experimental results and comparison. Finally,
section 4 concludes the paper.
2. TWOFISH ALGORITHM AND
RELATED WORKS
Twofish is a symmetric key block cipher with a block size of
128 bits and key sizes up to 256 bits, and it is related to the
earlier block cipher Blowfish [13] [14]. The cipher is a 16-
round Feistel network with a bijective F function made up of
four key- dependent 8-by-8-bit S-boxes, a fixed 4-by-4
maximum distance separable matrix over GF(28), a pseudo-
Hadamard transform, bitwise rotations, and a carefully
designed key schedule [15] [16]. Twofish can be easily
explained with the diagram as shown in Figure 1; 128-bit
plain-text (divided into four parts of 32-bit each) is given for
the input whitening where it is XOR-ed with four keys then
function g PHT which are explained under the heading
twofish functions and modules. Twofish can be implemented
in hardware in 14000 gates [10] [17]. The design of both the
round function and the key schedule permits a wide variety of
tradeoffs between speed, software size, key setup time, gate
count, and memory. Twofish has been extensively
cryptanalyzed used; even the best attack is able to break only
five rounds of the algorithm. Twofish designed with 16-round
Feistel network with a bijective F function. The main idea of
the round is repeated iteration and generate strong encryption
algorithm [10] [18] [19].

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
2
Figure 1: Twofish algorithm structure [20]
Twofish algorithm used in many research areas for getting the
best results for securing data. In [2] researcher used agile
methods of five phases and implements it using Chilkat
library. The data encrypted and decrypted permanently. [1]
Extend new cipher algorithm derived from Twofish called
Twofish-Ext256. It is same as Twofish algorithm with adding
some XOR operations in its design and 10 round Feistel
network. The result of comparing between to algorithm shows
that Twofish-Ext256 faster than Twofish. [8] The authors
used two algorithms Twofish and steganographic algorithm
first one for provides the speed of the system and the security,
the second algorithm for hiding the encrypted data into an
image. The new mixed algorithm coves the security and
efficiency.
3. IMPLEMENTATION
A. Environment
Towfish algorithm implemented using C language that widely
used in implementation of algorithms. In this paper we use
twofish encryption algorithm as a sequential source code and
modified it to run in parallel environment using MPI library.
We use three key sizes of algorithm 128, 192 and 256 bit for
the same text for both executions sequential and parallel code.
Sequential test executed on personal computer his
specification described in table 1.
Table 1: The Hardware and Software Specifications for
Computer used for Sequential execution.
Feature
Specification
Processor Clock
Intel(R) Core (TM) i5-3337U
CPU.
Memory
4096 MB RAM
Speed
@1.80GHz (4 CPUs).
Operating System
Windows 10 Enterprise 64-bit
File size
35 KB, 80 KB, 260 KB, 550 KB
Number of
Processors
1, 2, 4, 8, 16
The same seniors run in a sequential execution also run in
parallel code with changing numbers of CPUs used. This
experiment done based on different text sizes with three
lengths of keys 128, 192 and 256 bits. This testing focuses on
the effects of key length size and text size on encryption time.
The result of the sequential test compared with the result on a
parallel IMAN1 supercomputer.
IMAN1 Zaina cluster user to execute our parallel twofish
experiments using open MPI library. Supercomputer hardware
and software Specifications listed in table 2.
Table 2: The Hardware and Software Specifications for
IMAN1 Zaina cluster
Feature
Specification
Processor Clock
Intel(R) Core (TM) I5-6200U
CPU
Memory
8.00 GB RAM
Speed
@2.40 GHz
Operating System
Scientific Linux 6.4 with open
MPI 1.5.4, C and C++ compiler
File size
35 KB, 80 KB, 260 KB, 550 KB
Number of
Processors
1, 2, 4, 8, 16

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
3
B. Encryption process in Twofish
Encryption process convert message (plaintext) to unreadable
text (ciphertext). The test result using algorithm:
A. Insert plaintext:
Adding message test that went to encrypt in our test
input “HI..Jordan First”. That entered to algorithm
code.
B. Input key (hexa):
344646443145353742344339413634333842353337
39343235364242394f4552
C. Execute the algorithm and get the result
(chipertext):
60c68d37ed05c881327c2a046bfd3764153c580f60c
68d37ed05c881327c2a046bfd376460c68d37ed05c8
81327c2a046bfd376460c68d37ed05c881327c2a046
bfd3764153c580f064f4f04bb06e4eee3ff211b
C. Decryption process in Twofish
Decryption process returns the chipertext to the original
message using the key:
A. Insert chipertext:
60c68d37ed05c881327c2a046bfd3764153c580f60c
68d37ed05c881327c2a046bfd376460c68d37ed05c8
81327c2a046bfd376460c68d37ed05c881327c2a046
bfd3764153c580f064f4f04bb06e4eee3ff211b
B. Input key (hexa):
344646443145353742344339413634333842353337
39343235364242394f4552
C. Plaintext : return ciphertext to the original message
“HI..Jordan First”
4. EXPERIMENTAL RESULTS OF
TEOFISH ALGORITHM
The parallel twofish algorithm is implemented using open
Message Passing Interface (MPI) library, and executed on
IMAN1 supercomputer. MPI library provides different
functions to support distributing data among different
processors to be processed simultaneously. The MPI_Scatterv
procedure is used to split and distribute the input data on
processors. Every processor executes the same code with the
same key for all concurrently on data portion which allocated
to it. Parallel twofish evaluated by multiple input sizes (35,
80, 260, and 550) and key sizes of 128, 192, and 256 bits on
different number of processors (2, 4, 8, and 16). Figure 3, 4,
and 5 shows the execution time of the twofish encryption
algorithm on different number of processors with different
key lengths.
Figure 2: The execution time of twofish algorithm in one
processor.
Figure 2 show the execution time of Twofish encryption
algorithm on single processor with various packet sizes. The
analyses of execution time for execute sequential Twofish
implementation on one processor with different plaintext size
described in table 1. By comparing execution time to
encryption 2670.34(s) for plaintext with size 35 kb using 128
key length with 4056(s) execution time for same text size with
256 key length we see that execution time decreases when we
using larger key length. The same behavior appears by
increasing the plaintext size as table 1 shows for all
experiment plaintext in this case study (35kb, 80 kb, 260 kb
and 550kb). In plaintext 550kb the execution time decreases
around 3583s when we change key length from 128 to 256.
Also when key length changed from 128 to 256 in plaintext
size 260 kb the execution time decreases 2246s. According to
these results we compared the sequential execution and the
parallel execution in this paper for all plaintext size and keys.
Table 3: Experimental analysis of the sequential execution
Twofish algorithm with various packet sizes.
key
length
Execution time on 1 CPU with various packet
sizes.
35 kb
80 kb
260 kb
550 kb
128
2670.34
4620.11
5631.01
10631
192
3410
4751.2
6620.21
11548.6
256
4056
5936.12
7877.63
14214.9
Figure 3: Execution time of twofish - key length 128
0
5000
10000
15000
35
80
260
550
Time (S)
Text size in kb
key length 128
key length 192
key length 256
0
10000
20000
1
CPU
2
CPU
4
CPU
8
CPU
16
CPU
Time (s)
Number of Processors
550
260
80
35

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
4
Figure 4: Execution time of twofish - key length 192
Figure 5: Execution time of twofish - key length 256
In our experiments, we fixed key length size (128, 192, and
256 bits), also fixed file sizes (35, 80, 260 and 550 kb). this is
implemented and run in five cases in different numbers of
CPUs. For first time the three key sizes run in 2 CPU in
variant file size and the experiment repeated with increasing
the number of CPUs as the figures 3, 4, 5 shows and Table 3.
The speedup is calculated by taking the ratio between the
serial and parallel time. According to the results in Figures 6,
7, 8, 9 the best case of speedup for the data size of 35,
80,260, and 550 KB when the number of processors is equal
to 8 and 16, and the speedup decreases when the number of
processors is more than 16. Table 5 shows the results for
cases in our experiments.
Figure 6: The speedup of the three Key Length and data
size 35 kb
Figure 7: The speedup of the three Key Length and data
size 80 kb
Figure 8: The speedup of the three Key Length and data
size 260 kb
Figure 9: The speedup of the three Key Length and data
size 550 kb
Parallel efficiency is computed by taking the ratio between
speedup and number of processors. The figures 10, 11, 12, 13
shows the parallel efficiency of twofish algorithm for different
packet sizes on different number of CPUs. The results
describe that the efficiency is decreased with increasing the
key lengths for evaluated packet sizes.
Table 6 present the performance comparisons for runtimes in
parallel execution for different numbers of CPUs. The results
shows in all cases (in different text size) by increasing number
of CPUs the performance decreased because the
communication overhead increased; where the benefit of
parallelism occurred until 8 CPUs. In 16 CPUs the result is
become deceased comparing with 8 CPUs. In text size 550 kb
the efficiency is 0.068 for all key lengths. But comparing with
8 CPUs we find the efficiency changed by changing key size.
0
10000
20000
30000
1
CPU
2
CPU
4
CPU
8
CPU
16
CPU
Time (s)
Number of Processors
550
260
80
35
0
10000
20000
30000
1
CPU
2
CPU
4
CPU
8
CPU
16
CPU
Time (s)
Number of Processors
550
260
80
35
0
0.5
1
1.5
2
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
Speedup
Number of Processors
128
192
256
0
0.5
1
1.5
2
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
Speedup
Number of Processors
128
192
256
0
0.5
1
1.5
2
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
Speedup
Number of Processors
128
192
256
0
0.5
1
1.5
2
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
Speedup
Number of Processors
128
192
256

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
5
Figure 10: The Efficiency of the Tree Key Length and data
size 35 kb
Figure 11: The Efficiency of the Tree Key Length and data
size 80 kb
Figure 12: The Efficiency of the Tree Key Length and data
size 260 kb
Figure 13: The Efficiency of the Tree Key Length and data
size 550 kb
By increasing the CUPs in our experiments and comparing the
efficiency of the result at each case. We find that the less of
efficiency appears when we start using 16 CPUs, but if we
used more than 16 CPU the tests are noteworthy. Table 6
appears the efficiency results for all cases in our experiment in
details.
Table 4: comparing results using different key length with different test size and number of processors
Seq.NO
key length
Number of Processors
Time (in second) taken on Different text Size in Kb
35
80
260
550
1
128
1
2670.34
4620.11
5631.01
10631.01
2
128
2
2201.3
4102.5
5118
9321.7
3
128
4
1956.04
3826.1
4863.54
8932.8
4
128
8
1736.2
3793.1
4801.2
7621.4
5
128
16
2570.34
3978
4986.35
9745.6
6
192
1
3410
4751.2
6620.21
11548.56
7
192
2
2631
4028.01
6238.04
9354.62
8
192
4
2121
3492.6
5823.7
8925.641
9
192
8
1863
2946.2
5688.3
7352.14
10
192
16
3410
4002
5634.64
10647.25
11
256
1
4056
5936.12
7877.63
14214.86
12
256
2
3562
4259.32
6954.02
10058.33
13
256
4
3263
4063
6235.47
9625.14
14
256
8
2564
3218
4963.41
9217.9
15
256
16
3968
5900
7529.78
13125.78
0
0.2
0.4
0.6
0.8
1
1.2
1
CPU
2
CPU
4
CPU
8
CPU
16
CPU
Efficiency
Number of Processors
key length 128
key length 192
key length 256
0
0.2
0.4
0.6
0.8
1
1.2
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
Efficiency
Number of Processors
128
192
256
0
0.2
0.4
0.6
0.8
1
1.2
1
CPU
2
CPU
4
CPU
8
CPU
16
CPU
Efficiency
Number of Processors
key length 128
key length 192
key length 256
0
0.2
0.4
0.6
0.8
1
1.2
1
CPU
2
CPU
4
CPU
8
CPU
16
CPU
Efficiency
Number of Processors
key length 128
key length 192
key length 256

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
6
Table 5: Speedup for Parallel Twofish algorithm with Different File sizes
key length
Speed up of the parallel execution on different number of processors
Text Size(k)
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
128
1
1.213
1.365
1.538
1.039
35
192
1
1.296
1.608
1.83
0.295
35
256
1
1.139
1.243
1.532
1.022
35
128
1
1.126
1.208
1.218
1.61
80
192
1
1.18
1.36
1.613
1.187
80
256
1
1.394
1.461
1.845
1.006
80
128
1
1.1
1.158
1.173
1.129
260
192
1
1.061
1.137
1.164
1.175
260
256
1
1.133
1.263
1.587
1.046
260
128
1
1.14
1.19
1.395
1.091
550
192
1
1.232
1.294
1.571
1.085
550
256
1
1.413
1.477
1.542
1.083
550
Table 6: Efficiency for Parallel Twofish algorithm with Different File sizes
key length
Efficiency for parallel execution on different number of processors
Text Size(k)
1 CPU
2 CPU
4 CPU
8 CPU
16 CPU
128
1
0.607
0.341
0.192
0.065
35
192
1
0.648
0.402
0.229
0.018
35
256
1
0.569
0.311
0.198
0.064
35
128
1
0.563
0.302
0.152
0.073
80
192
1
0.59
0.34
0.202
0.074
80
256
1
0.697
0.365
0.231
0.063
80
128
1
0.55
0.289
0.147
0.071
260
192
1
0.531
0.284
0.145
0.073
260
256
1
0.566
0.316
0.198
0.065
260
128
1
0.57
0.298
0.174
0.068
550
192
1
0.617
0.323
0.196
0.068
550
256
1
0.707
0.369
0.193
0.068
550
5. CONCLUSION
In this research, we display the performance of parallel
twofish algorithm that was evaluated according to execution
time, speedup and efficiency for different sizes of data and
various numbers of processors. Parallel twofish in this paper
was implemented by C++ using open MPI library and
executed on IMAN1 supercomputer. In parallel twofish,
According to results, parallel twofish has better execution
time for large date size than small data size but with large
number of processors on small data size will increase running
time instead of decrease execution time because amount of
communication between processors will be huge. The
experimental results show that the running time will be
decreased, and the speed-up of encryption and decryption
processes will be increased when the number of processors is
8. Future work for this paper is encryption and decoding of
huge measure of content records, pictures, sound records and
video files also large text files.
6. REFERENCES
[1] M. Saadeh, H. Saadeh and M. Qatawneh, "Performance
Evaluation of Parallel Sorting Algorithms on IMAN1
Supercomputer," 2016.
[2] M. Qatawneh , "embedding linear array network into the
tree-hypercube network," 2005.
[3] M. Qatawneh, "Multilayer Hex-Cells: A New Class of
Hex-Cell Interconnection Networks for Massively
Parallel Systems," 2011.
[4] M. Qatawneh, "Embedding Binary Tree and Bus into
Hex-Cell Interconnection Network," 2011.
[5] M. Qatawneh and H. Khattab, "New Routing Algorithm
for Hex-Cell Network," 2015.

International Journal of Computer Applications (0975 – 8887)
Volume 179 – No.50, June 2018
7
[6] M. Qatawneh, "New Efficient Algorithm for Mapping
Linear Array into Hex-Cell Network," 2016.
[7] Aparna, A. K, J. Solomon, H. M and I. V, "A Study of
Twofish Algorithm," 2016.
[8] M. S. Saraireh, "A Novel Security Scheme based on
Twofish and Discrete Wavelet Transform," 2017.
[9] M. Ebrahim, S. Khan and U. Bin Khalid, "Symmetric
Algorithm Survey: A Comparative Analysis," 2013.
[10] R. KAUR and J. SINGH SAINI, "extended twofish block
cipher algorithm for 256-bit blocks and performance
comparison with twofish," 2014.
[11] P. Chodowiec, P. Khuon and K. Gaj, "Fast
Implementations of Secret-Key Block Ciphers Using
Mixed Inner- and Outer-Round Pipelining," 2001.
[12] Mushtaque, H. Dhiman, S. Hussain and S. Maheshwari,
"Evaluation of DES, TDES, AES, Blowfish and Two fish
Encryption Algorithm: Based on Space Complexity,"
2014.
[13] H. K. Verma and R. K. Singh, "Performance Analysis of
RC6, Twofish and Rijndael Block Cipher Algorithms,"
2012.
[14] P. Gehlot, R. Sharma and S. R. Biradar, "VHDL
Implementation of Twofish Algorithm".
[15] D. Rane, "Superiority of Twofish over Blowfish," 2016.
[16] S. Lucks, "The Saturation Attack – A Bait for Twofish,"
2002.
[17] B. Schneier, J. Kelsey, D. Whiting, D. Wagner, C. Hall
and N. Ferguson, "On the Twofish Key Schedule," 1999.
[18] M. Dener, "Security Analysis in Wireless Sensor
Networks," 2014.
[19] P. Gehlot, S. R. Biradar and B. P. Singh,
"Implementation of Modified Twofish Algorithm using
128 and 192-bit keys on VHDL," 2013.
[20] M. A. Devi and M. R. B. S, "Two fish Algorithm
Implementation for lab to provide data security with
predictive analysis," 2017.
IJCATM : www.ijcaonline.org