Page 61 - Fister jr., Iztok, and Andrej Brodnik (eds.). StuCoSReC. Proceedings of the 2018 5th Student Computer Science Research Conference. Koper: University of Primorska Press, 2018
P. 61
main tool for working on the experiment is the Intel R Table 3: Display of communication between parties
Core TM i7-5500U processor, 8GB RAM and the Windows
10 Pro version 1709. Post-quantum asymmetric encryption and security bits
algorithms were implemented in the Libqos library, written
in C, as part of the OQS project. Modern asymmetric en- Name A → B (bytes) B → A (bytes)
cryption algorithms are implemented within the Crypto++
open source library, written in C++ language. We installed Frodo 11377 11296
these libraries for the Visual Studio 2017 development envi-
ronment, on Windows SDK version 10.0.17134.0, on a 64bit BCNS15 4096 4096
platform. Once all tests where performed, the data was ag-
gregated and analyzed in the Microsoft Excel tool. Graph- MSR LN16 1824 2048
Pad Prism tool was used to draw graphs that will show the
results of the experiment. NewHope 1824 2048
4.2 Implementation of the experiment SIDH p503 378 378
To perform an experiment, i.e., measurements of the effi- SIDH p751 564 564
ciency of individual asymmetric modern and post-quantum
cryptographic algorithms, two test cases had to be performed: SIKE p503 378 402
One test case for measuring modern asymmetric encryption
algorithms and the other for measuring post-quantum asym- SIKE p751 564 596
metric encryption algorithms. Modern asymmetric encryp-
tion algorithms were used from the Crypto++ library ver- DH-1024 128 128
sion 7.0.0. Test case, the console application, was built in
Release Mode on a 64bit platform. The algorithms selected DH-512 64 64
for measurement were: RSA (1024, 2048, 3072), DH (512,
1024), DH (1024) with predefined parameters, ECC256. Each ECDH256 32 32
algorithm was running over a time period of 30 seconds,
where we measured the number of iterations. The number RSA-1024 128 128
of runs for each algorithm was saved in the .csv file, together
with the name and spent time. The test case was limited to RSA-2048 256 256
11 repetitions of measurement of each algorithm, each of the
7 algorithms was run 11 times in 30 seconds. The test case RSA-3072 384 384
was performed four times. Each lasted about 40 minutes,
totalling about 2 hours and a half. For RSA, the number of Table 4: Display of classical and quantum security
key generation and validation were measured, while for DH
and EEC the number of generations and key aggremments. bits
Post-quantum asymmetric encryption algorithms were used Name Classical bits Quantum bits
from the libqos library, which is part of the OQS prototype
project. Test case, the console application was generated in Frodo 144 130
Release Mode on a 64bit platform. The test case is sim-
ilar to the test case for modern cryptographic encryption BCNS15 86 87
algorithms. The algorithms selected for measurement were:
LWE Frodo recommended, RLWE BCNS15, RLWE MSR MSR LN16 128 112
LN16, RLWE NewHope, SIDH MSR p503 and SIDH MSR
p751, SIKE MSR p503 and SIKE MSR p751. For each algo- NewHope 229 206
rithm, the number of implementations over a time period of
30 seconds was measured. The number of runs for each algo- SIDH p503 192 128
rithm was saved in the .csv file, together with the name and
spent time. The test case was limited to 11 repetitions of SIDH p751 256 192
measurement of each algorithm, each of the algorithms was
run 11 times in 30 seconds. The test case was performed SIKE p503 192 128
four times. Each lasted about 4 minutes, totalling about 3
hours. For all algorithms the number of generation and key SIKE p751 256 192
agreements were measured.
DH-1024 73 -
4.3 Results and data analysis
DH-512 50 -
Apart from reliable cryptographic properties, the ideal algo-
rithm for asymmetric encryption should have as little com- ECDH256 136 -
munication requirements as possible, low memory consump-
tion and the least possible execution time. This paper takes RSA-1024 73 -
into account the speed of execution and communication re-
quirements as the main metrics for comparison, while the RSA-2048 103 -
memory consumption is ignored.
RSA-3072 128 -
4.3.1 Communication requirements
The communication between the parties, A → B and B →
A, was obtained from the program code and the implemen-
tation of algorithms which was verified with data in sources
[10] [12] [3] [4] [1] [5] [6] [9]. The table 3 shows the commu-
nication requirements between parties, the amount of data
exchanged by the parties when the keys were generated and
the table 4 shows the classical and quantum security bits of
the algorithm.
4.3.2 Time complexity
Figure 1 shows the time required for the tested algorithm
is executed once in milliseconds. From the graph, it can be
determined which algorithms are more successful i.e., faster.
4.3.3 Interpretation of results
The most effective post-quantum asymmetric encryption al-
gorithms are BCNS15, LN16 and NewHope. LN16 and
NewHope were faster than all post-quantum and also of all
classic asymmetric encryption algorithms. The disadvan-
tage of the fastest post-quantum algorithms is relatively high
communication complexity. In generating and distributing
the keys, the LN16 and NewHope algorithms make 3,872
StuCoSReC Proceedings of the 2018 5th Student Computer Science Research Conference 63
Ljubljana, Slovenia, 9 October
Core TM i7-5500U processor, 8GB RAM and the Windows
10 Pro version 1709. Post-quantum asymmetric encryption and security bits
algorithms were implemented in the Libqos library, written
in C, as part of the OQS project. Modern asymmetric en- Name A → B (bytes) B → A (bytes)
cryption algorithms are implemented within the Crypto++
open source library, written in C++ language. We installed Frodo 11377 11296
these libraries for the Visual Studio 2017 development envi-
ronment, on Windows SDK version 10.0.17134.0, on a 64bit BCNS15 4096 4096
platform. Once all tests where performed, the data was ag-
gregated and analyzed in the Microsoft Excel tool. Graph- MSR LN16 1824 2048
Pad Prism tool was used to draw graphs that will show the
results of the experiment. NewHope 1824 2048
4.2 Implementation of the experiment SIDH p503 378 378
To perform an experiment, i.e., measurements of the effi- SIDH p751 564 564
ciency of individual asymmetric modern and post-quantum
cryptographic algorithms, two test cases had to be performed: SIKE p503 378 402
One test case for measuring modern asymmetric encryption
algorithms and the other for measuring post-quantum asym- SIKE p751 564 596
metric encryption algorithms. Modern asymmetric encryp-
tion algorithms were used from the Crypto++ library ver- DH-1024 128 128
sion 7.0.0. Test case, the console application, was built in
Release Mode on a 64bit platform. The algorithms selected DH-512 64 64
for measurement were: RSA (1024, 2048, 3072), DH (512,
1024), DH (1024) with predefined parameters, ECC256. Each ECDH256 32 32
algorithm was running over a time period of 30 seconds,
where we measured the number of iterations. The number RSA-1024 128 128
of runs for each algorithm was saved in the .csv file, together
with the name and spent time. The test case was limited to RSA-2048 256 256
11 repetitions of measurement of each algorithm, each of the
7 algorithms was run 11 times in 30 seconds. The test case RSA-3072 384 384
was performed four times. Each lasted about 40 minutes,
totalling about 2 hours and a half. For RSA, the number of Table 4: Display of classical and quantum security
key generation and validation were measured, while for DH
and EEC the number of generations and key aggremments. bits
Post-quantum asymmetric encryption algorithms were used Name Classical bits Quantum bits
from the libqos library, which is part of the OQS prototype
project. Test case, the console application was generated in Frodo 144 130
Release Mode on a 64bit platform. The test case is sim-
ilar to the test case for modern cryptographic encryption BCNS15 86 87
algorithms. The algorithms selected for measurement were:
LWE Frodo recommended, RLWE BCNS15, RLWE MSR MSR LN16 128 112
LN16, RLWE NewHope, SIDH MSR p503 and SIDH MSR
p751, SIKE MSR p503 and SIKE MSR p751. For each algo- NewHope 229 206
rithm, the number of implementations over a time period of
30 seconds was measured. The number of runs for each algo- SIDH p503 192 128
rithm was saved in the .csv file, together with the name and
spent time. The test case was limited to 11 repetitions of SIDH p751 256 192
measurement of each algorithm, each of the algorithms was
run 11 times in 30 seconds. The test case was performed SIKE p503 192 128
four times. Each lasted about 4 minutes, totalling about 3
hours. For all algorithms the number of generation and key SIKE p751 256 192
agreements were measured.
DH-1024 73 -
4.3 Results and data analysis
DH-512 50 -
Apart from reliable cryptographic properties, the ideal algo-
rithm for asymmetric encryption should have as little com- ECDH256 136 -
munication requirements as possible, low memory consump-
tion and the least possible execution time. This paper takes RSA-1024 73 -
into account the speed of execution and communication re-
quirements as the main metrics for comparison, while the RSA-2048 103 -
memory consumption is ignored.
RSA-3072 128 -
4.3.1 Communication requirements
The communication between the parties, A → B and B →
A, was obtained from the program code and the implemen-
tation of algorithms which was verified with data in sources
[10] [12] [3] [4] [1] [5] [6] [9]. The table 3 shows the commu-
nication requirements between parties, the amount of data
exchanged by the parties when the keys were generated and
the table 4 shows the classical and quantum security bits of
the algorithm.
4.3.2 Time complexity
Figure 1 shows the time required for the tested algorithm
is executed once in milliseconds. From the graph, it can be
determined which algorithms are more successful i.e., faster.
4.3.3 Interpretation of results
The most effective post-quantum asymmetric encryption al-
gorithms are BCNS15, LN16 and NewHope. LN16 and
NewHope were faster than all post-quantum and also of all
classic asymmetric encryption algorithms. The disadvan-
tage of the fastest post-quantum algorithms is relatively high
communication complexity. In generating and distributing
the keys, the LN16 and NewHope algorithms make 3,872
StuCoSReC Proceedings of the 2018 5th Student Computer Science Research Conference 63
Ljubljana, Slovenia, 9 October
