Professional Documents
Culture Documents
Hardware Implementation
for Fast Encryption
이 훈 재 (李 焄 宰) Hoon-Jae Lee
CNSL
Cryptography and Network Security Lab.
hjlee@dongseo.ac.kr
http://kowon.dongseo.ac.kr/~hjlee
http://crypto.dongseo.ac.kr
2013-05-09 CNSL-Internet-DongseoUniv. 1
Agenda
I. Overview of hardware implementation
II. Hardware Design Approaches
2.1 Hardware design approaches
2.2 ASIC/FPGA
2.3 Pipeline/Parallel-structures
1
High-Speed Implementation
- Memory Access Time (ns)
Samsung DDR SRAM Speed (2004.7)
2013-05-09 CNSL-Internet-DongseoUniv. 3
2013-05-09 CNSL-Internet-DongseoUniv. 4
2
- Adder & Shifter Speed (ns)
2013-05-09 CNSL-Internet-DongseoUniv. 5
2013-05-09 CNSL-Internet-DongseoUniv. 6
3
High-Speed Implementation
Algorithm XOR
– Core Operation
Mod 232 Mod 232 Fixed Variable Mod 232 GF(28) S-box
Add sub shift rotate Mul Mul *LUT
*LUT
Notations >>>
<<<
Max. Sever 180 ps 180 ps Several Several 12.8 ns Several Several
Speed(ASI al 10 10 ps 10 ps ns ns
C) ps
MARS ● ● ● ● ● ● ●
RC6 ● ● ● ● ●
Rijndael ● ● ● ●
Serpent ● ● ●
Twofish ● ● ● ● ●
2013-05-09 CNSL-Internet-DongseoUniv. 7
4
1.1 AES evaluation - Initial(1997, NIST)
5
1.1 AES evaluation - Final(2000, NIST)
qGovernment à ASIC
ØNSA
ØIBM, Inc.
ØMisubishi, Inc.
qUniversities à FPGA
ØUC Berkeley
ØUSC(University of Southern California)
ØWPI(Worcester Polytechnic Institute )
ØGMU(George Mason University)
ØMicronic, Inc.
6
1.2 Somethings to design approaches
q Crypto/Non-Crypto Module
Ø Crypto Module: Refer to FIPS 140-1,2,3 (CMVP)
Ø Non-Crypto Module: Power module, Interface module, etc.
Implementation Platforms
14
7
1.2 Somethings to design approaches
Which Platform?
q The choice of the implementation platform is driven by a
multitude of factors, which include,
Ø Performance needed
Ø Development and per-unit costs
Ø Power consumption (Important in case of wireless devices!)
Ø Flexibility
Ø Physical Security
15
Platform Characteristics
16
8
1.2 Somethings to design approaches
qWhy Reconfigurable? …
17
Ø Algorithm agility
Ø Algorithm Upgradability
Ø Architecture Efficiency
Ø Resource Efficiency
Ø Algorithm Modification
Ø Throughput (Relative to software)
Ø Cost Efficiency (Relative to ASICs)
18
9
1.2 Somethings to design approaches
19
Ø State Register
Ø Key Register
Ø Updating Logic
Ø Control & Load Logic
20
10
1.2 Somethings to design approaches
21
11
1.2 Somethings to design approaches
q Examples
u Project 25 OTAR Documents : TIA/EIA APCO Project 25
Non-Crypto M Non-Crypto M
12
II. Hardware Design Approaches
q Design Tools
Ø FPGA(Field Programmable Gate Array)
Ø ASIC(Application Specific Integrated Circuits)
Ø NT(nano-technology)
q Design approaches
Ø Round-Common: implementation for one-round function
à lower gates, lower costs, but low-speed
Ø Pipeline-structured: implementation for full-round sequential functions
à higher gates, higher costs, but high-speed
Ø Parallel : parallel operation for round-common or pipeline-structured
Ø Parallel & Pipeline : combinations of parallel and pipeline-structured
13
2.2 ASIC vs. FPGA
Items ASIC FPGA
14
2.2 FPGA Device - example(Target)
q Design House
Ø Samsung (http://www.samsungelectronics.com/semiconductors/asic/ASIC.htm)
Ø Hynix (http://www.hynix.com/)
Ø CNC technology (http://www.cnstec.com/)
Ø ADC (http://www.adc.co.kr/)
Ø TLI (http://www.tli.co.kr/)
Ø ECT (http://www.ect.co.kr/)
Ø INC technology (http://www.inctech.co.kr/)
Ø ARALION (http://www.aralion.com)
15
2.3 Pipeline-structured
q Input – (unit 1 … unit n) - Output
2.3 Pipeline-structured
q Throughputs (s : # of input data, n : # of pipeline steps)
T1 s´n
Sp = =
Tn n + ( s - 1)
q Efficiency:
Sp
E= ´ 100
n
q Max. throughputs:
n
lim S p = lim =n
s ®¥ s ®¥ n 1
+ (1 - )
s s
16
2.3 pipeline vs. round-common
라운드 공유 방식 라운드 공유 방식
라운드 구성에 따른 결과 출력도 라운드 방식에 따른 게이트 소요도
파이프라인 방식 파이프라인 방식
1 만 단위 총 소요 게이트 수
1152 78
72
총 결과 출력 비트
1024 66
896 60
768 54
48
640 42
512 36
30
384 24
256 18
12
128 6
0
0
1 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
라운드 수행 횟수 라운드 소요 횟수
17
Block Ciphers: Key Elements
q Bitwise XOR, AND, OR.
q Addition or subtraction modulo 2n
q Shift or rotation by a constant number of bits.
q Data-dependent rotation by a variable number
of bits.
q Multiplication modulo the table entry value.
q Multiplication in the Galois field specified by the
table entry value.
q Inversion modulo the table entry value.
q Look-up-table substitution
18
3.1 High-speed core components
- Memory Access Time (ns)
삼성전자 DDR SRAM Speed (2004.7월)
19
3.1 High-speed core components
- Adder & Shifter Speed (ns)
q 32-bit Adder Speed[SAMSUNG 0.13um ASIC] à 387ps
l 12.8ns @ 0.35um
20
3.1 High-speed core components
– Core Operation
Algorithm XOR Mod 232 Mod 232 Fixed Variable Mod 264 Mul GF(28) S-box
Add sub shift rotate Mul *LUT
*LUT
Notation >>>
<<<
MARS ● ● ● ● ● ● ●
RC6 ● ● ● ● ●
Rijndael ● ● ● ●
Serpent ● ● ●
Twofish ● ● ● ● ●
21
3.2 Implementation Examples in Block cipher
- AES: George Mason Univ’2000
22
3.2 Implementation Examples in Block cipher
- AES: George Mason Univ’2000
23
3.2 Implementation Examples in Block cipher
- AES pipeline [2002-2006]
Designed ASIC or Performan # Integrated Published Remarks
by FPGA ce[Gbps] [Gates] to
Lee, ASIC 2.56 CISC’200
YoonKyun 0.25um 2
g et.al
Goo, FPGA 20 WISC’200 40-
BonSeog 2 stage(10)
et.al Pipeline
McLoone, FPGA 12.02 IEEE
McCanny Syp2000
Hodjat FPGA 21.54 5177 FCCM200
et.al CLBslices 4
/UCLA
Saggese FPGA 20.3 5810 FPL2003
et.al CLBslices
Goo, FPGA Under 3,992 KIISC RFID for
BonSeog 0.25um Gbps Journal low-power
et.al 2006.10
24
3.2 Implementation Examples in Block cipher
- ARIA
Designed ASIC or Performan # Integrated Publishe Remarks
by FPGA ce[Gbps] [Gates] d to
NSRI ASIC 1,871 8,935 WISC200 Non-pipeline
Team Hynix 3 [AES]*
0.25um [1,839]* [9,088]* High-speed,
low-
integrated
Park, FPGA 1,142 29,930 ga WISC200 Non-pipeline
JinSeob 0.22um 1,599 sl 4
et.al
Yoo, Xilinx 437 1,490 slices IEEK, Non-pipeline
YeongGab VertexE- 2005.4
et.al 1600 FPGA
Jang, 64-bit - - KIISC -
HwanSug microproc Journal
et.al essor 2006.6
c) Round function
25
3.2 Implementation steps– IDEA= 7 steps
- 8-round transformation
- 8-round transformation
(6 x 8 = 48 subkeys)
(6 x 8 = 48 subkeys)
- output transformation
- output transformation
(4 subkeys)
(4 subkeys)
- Key generation
- Key generation
(16-bit, 52 subkeys)
(16-bit, 52 subkeys)
- Multiplication-Addition
- Multiplication-Addition
(MA) structures
(MA) structures
a) Block Diagram
b) Round functions
26
3.2 Implementation steps– ARIA= 3 steps
128/192/ 128/192/2
Key size 56 112 128 128
256 56
# rounds 16 16 x 3 8 12/14/16 16 10/12/14
128x13/ 128x11/
Roundkey 48x16
48x16 16x52 128x15/ 32x2x16 128x13/
size ,3
128x17 128x15
8x8, 4
S-BOX 6x4, 8 6x4, 8 16-bit MA 8x8, 2 8x8
8x32, 4
Key Space 256 2112 2128 2128~256 2128 2128~256
Proper CPU
8-bit 8-bit 16-bit 32-bit 32-bit 32-bit
(word size)
Year 1975-77 1979 1990-92 2003 1997 2000-01
Switzerlan
Country USA USA KOREA KOREA USA
d
TTAS.KO
Standard FIPS-46 - - - FIPS-197
-12.0004
27
3.2 Implementation steps– comparisons
Cipher Clock #of round # Steps in Speed
Period(ns) (#of clock) 1-round (Mbps)
28
Conclusion (On Designing Block cipher)
Approaches Descriptions Properties
1) HW Design FPGA(Field Programmable Gate Array) - Low cost, easy design at Lab./research Inst.
Tools * Xilinx Vertex/Altera FLEX, etc - Lower speed to ASIC if the same layout
ASIC(application specific IC) - High cost, difficulty in design at design house
* Samsung/Hynix/Lucent Tech. etc - Higher speed to FPGA about 5~10 times if the
same layout
2) Semi- 0.5μm → 0.35 μm → 0.22 μm → 0.18 μm Decreasing layout → High speed, low power,
conductor layout → 0.13 μm (→ 90nm → 70nm → 60nm → 50nm) Down sizing
3) Pipeline Round-common →◘ → Single round chip (1)
approaches
Pipeline-structured →◘◘◘◘◘→ multiple k-round chip (k times in speed)
* partial pipeline (k=#round, #round/divisor,
* full pipeline or #round x positive integer)
Parallel-structured →◘ → Single round chip x n parallel (n times)
→◘ →
Parallel Pipeline →◘◘◘◘◘→ multiple k-round chip x n parallel (n x k times)
→◘◘◘◘◘→
4) Component Selection of high-speed components High-speed component
approaches Ex) XOR, Mod232adder, MUL, SFT, S-P (Security needs in the first important)
29
4.1 Higher speed of LFSR
- LFSR-based stream cipher
0 1 2 3 4 Output
System clock
5 3 2
- Primitive polinomial : P ( x) = x + x + x + x + 1
Q D Q .D. . . Q D Q D Q D Output
0 1 2 3 4
30
4.1 Higher speed of LFSR
- LFSR-based stream cipher
Ì Parallel Stream Cipher
q Basic components in parallel stream cipher
Ø PS-LFSR, Serial/Parallel Converter
Ø m-Parallel Nonlinear Functions
q Parallel-Shifting LFSR(PS-LFSR)
Ø Crypto-degree [ Similar to LFSR
Ø Ext. clock [ m-bit parallel output/clocking [ m times faster
( 2£m£n )
----------------------------------------
[Ref] HJ Lee, SJ Moon” Parallel Stream Cipher for Secure High-Speed Communications”
Signal Processing, Vol.82, No. 2, Feb. 2002.
SYSTEM CLOCK
a) n-stage LFSR
feedback m
Feedback connection m : m-bit left-shifting of the original combination
feedback 2
Feedback connection 2 : 1-bit left-shifting of the original combination
-1 0 1 ...... n-2
feedback 1
Feedback connection 1: the original combination
. . .
0 1 2 ...... n-1
SYSTEM CLOCK
b) (n,m) PS-LFSR
31
4.1 Higher speed of LFSR
SYSTEM CLOCK
feedback 8
Feedback connection 8 : s(33+t)= s(28+t) ^ s(-5+t) ^ s(-6+t) ^ s(-7+t)
-7 -6 -5 28
. . . . . .
feedback 2
Feedback connection 2 : s(39+t)= s(34+t) ^ s(1+t) ^ s(t) ^ s(-1+t)
-1 0 1 34
feedback 1
Feedback connection 1: s(40+t)= s(35+t) ^ s(2+t) ^ s(1+t) ^ s(t)
. . .
0 1 2 35
-8 7 15 23 31 39
-7 6 14 22 30 38
-6 5 13 21 29 37
. . . m=8 bits
-5 4 12 20 28 36
-4 3 11 19 27 35 OUTPUT
-3 2 10 18 26 34
-2 1 9 17 25 33
-1 0 8 16 24 32
SYSTEM CLOCK
b) (n,m)=(40,8) PS-LFSR
-7 -6 -5 30
......
feedback 2
Feedback connection 2
-1 0 1 37
feedback 1
Feedback connection 1
...
0 1 2 38
--- 7 15 23 31 ---
-7 6 14 22 30 38
-6 5 13 21 29 37
...
-5 4 12 20 28 36
m=8 bits
-4 3 11 19 27 35
OUTPUT
-3 2 10 18 26 34
-2 1 9 17 25 33
-1 0 8 16 24 32
SYSTEM CLOCK
p( x) = x 39 + x 37 + x 25 + x 24 + x 22 + x 8 + x 6 + x 4 + 1
b) (n,m)=(39,8) PS-LFSR
32
4.1 Higher speed of LFSR
C1 C2 ... CN C1 C2 ... CN
(x1) (x1)
l1-LFSR 1 Nonlinear l1-LFSR 1 Nonlinear
... ...
combiners: combiners:
(x2) (x2)
l2-LFSR 2 l2-LFSR 2
(memoryless (memoryless
(x3) or not) (x3) or not)
l3-LFSR 3 l3-LFSR 3
...... ......
(xN) (xN)
lN-LFSR N f 1, f 2, ..., f m lN-LFSR N f 1, f 2, ..., f m
m m
Serial/ Parallel/ Serial/ Parallel/
m m m m
Plaintext Parallel Serial Parallel Serial Plaintext
Converter Converter Converter Converter
Ciphertext
(xm1)
m
l2-PS-LFSR2
. . . . . . . . . . . . . . .
c m1 , c m 2 ,.., c mM
M M
(x1m)
Summation
(x2m) Generator:
(ym)
. . . SUM-BSGm
m
lm-PS-LFSRm
(xmm)
Note: N = m, M1 = M2 = ..= Mm = M
33
4.1 Higher speed of LFSR
feedback connection 4
feedback connection 3
feedback connection 2
00
0 01
02
Y 127-LFSR
39-stage LFSRca 03 Y 129-LFSR
89-stage 4-bit parallel bLFSR d
KEYDATA 1
SEL 1
k n
SEL CLK
... ...
KEYLDEN
fac(t)
(t) fb (t)z(t)
ffac fdb
CLOCK-CONTROL
CLOCK-CONTROL DATA GENERATION
CLOCK-CONTROL
SYSTEM CLOCK
KEY_LDEN
carry = S1
From FA sum = S0
feedback 4
...... feedback 2
feedback 1
S0 S1 S0 S1 S0 S1 S0 S1 S0 S1 S0 S1
d[88+t]
d[t] d[6+t]d[9+t] d[34+t]d[36+t]d[47+t] d[50+t]
s0 s1 s0 s1 s0 s1 s0 s1 s0 s1 s s0 s1
......
s0 s1
0 0 ...... 0 0 0 0 0
d-3 y 1 d0 y 1 d84 d85 y 1 d86 y 1 d87 y 1 d88 0 y 1
y 1
2 2 2 2 2 2 y 2
3 3 3 3 3 3 1 3
......
......
3-stage ......
LBUF(d-3,d-2,d-1)
SYSTEM_CLOCK
d0
d1 fd function output sequences
d2 KEYSTREAM
......
* filtered function
d88
34
4.1 Higher speed of LFSR
-7 -6 -5 30
......
feedback 2
Feedback connection 2
-1 0 1 37
feedback 1
Feedback connection 1
...
0 1 2 38
--- 7 15 23 31 ---
-7 6 14 22 30 38
-6 5 13 21 29 37
...
-5 4 12 20 28 36
m=8 bits
-4 3 11 19 27 35
OUTPUT
-3 2 10 18 26 34
-2 1 9 17 25 33
-1 0 8 16 24 32
SYSTEM CLOCK
p( x) = x 39 + x 37 + x 25 + x 24 + x 22 + x 8 + x 6 + x 4 + 1
b) (n,m)=(39,8) PS-LFSR
35
4.1 Higher speed of LFSR
36
4.1 Higher speed of LFSR
Ì LFSR vs. PS-LFSR
Items 39-LFSR (39,8) PS-LFSR
Period 2 39 - 1 2 39 - 1
Processing rate @ 500 x 8
500MHz system clock 500 Mbps = 4 Gbps
(Max. Delay = 1.73 ns)
37
4.2 Word Based Stream Ciphers
- Word Based LFSR
qWord Based LFSR
LFSR
q Examples:
l Snow
l Sober
l Turing
l Dragon Nonlinear Filter
Output
38
4.2 Word Based Stream Ciphers
- Dragon
q Dragon
Ø Word –based stream cipher : ICISC’2004
Ø Throughput: 23Gbps (H/W), 3.8Gbps (S/W)
Ø Cooperative results with Australia QUT-ISRC(Dr. Dawson)
Ø Submitted to the Int’l standard ECRYPT-eSTREAM (phase 3)
NLFSR M
FPDS Selection
feedback keystream
39
4.2 Word Based Stream Ciphers
- Dragon
vWord-based FSR (WFSR)
lSimilar to the bit-based LFSR (LFSR)
lFeedback shifted by the unit of the word-size (W)
40
4.2 Word Based Stream Ciphers
- Dragon
41
4.2 Word Based Stream Ciphers
- Dragon
42
4.2 Word Based Stream Ciphers
- Dragon
v (New) Dragon in Parallel-Structured (m=8)
43
4.2 Word Based Stream Ciphers
- Dragon
v Performance Analysis
Items Worst case Typical case Best case
memory 287,600 287,600 287,544
Area
comb. 8,126 8,068 8,219
(gate size)
total 295,726 295,688 295,763
Critical Path delay(ns) 14.36 10.26 6.72
Throughput 4.4 Gbps 6.2 Gbps 9.5 Gbps
Parallel-Throughput(m=8) 35.2 Gbps 49.6 Gbps 76 Gbps
(max. m=16) 70.4 Gbps 99.2 Gbps 152 Gbps
Conclusion
q Higher speed in block cipher
ØASIC latest tech.(layout) : 0.13um, 90–40nm
ØPipeline/Parallel pipeline-structured
ØThe smaller in #step of round, the faster
ØSelection of the faster components in step
(Security is the best required)
q Higher speed in stream cipher
ØParallel-Structured LFSR à m times faster
ØClock-controlled (compensate) à 1 time
ØWord-Based Stream Cipher à W times faster
44
V. Highly Reliable Synchronous
Stream Cipher System
Lee, HoonJae
이훈재 (李焄宰)
Division of Computer Information
Engineering,
Dongseo University, Busan, Korea
hjlee@dongseo.ac.kr
http://kowon.dongseo.ac.kr/~hjlee
I. Introduction
u Two kinds in Stream cipher :
Ø Self-synchronous Stream Cipher
u In this presentation,
Ø a highly reliable keystream synchronization method
45
2. Keystream Synchronization
In Synchrous Stream Cipher
2. Keystream Synchronization
In Synchrous Stream Cipher
46
2. Keystream Synchronization
In Synchrous Stream Cipher
2. Keystream Synchronization
47
Ex) OTAR Documents – TIA/EIA APCO Project 25
2. Keystream Synchronization
- TIA/EIA APCO Project 25 -
u APCO Project 25 - CAI Message Format
à NOT Highly-Reliable ENCRYPT-SYNC in Wireless Noisy Channel
(Window size N = 72 bit (small) !! è 128-bit required !!)
48
3. Stream Cipher System
n Encipering/Deciphering system
KEK
CODEC TEK
Cipher
CAI
Equipment
(Fig. 3)
(Fig. 2)
KMF
TEK
CODEC TEK
Cipher
CAI
Equipment
KMF (Fig. 3)
(Fig. 2)
TEK
KEK
(=128-bit SYNC)
98 Dongseo University,
Korea
49
3. Stream Cipher System
Technical Approach
q Stream Synchronization : Software Implementation
En En
Voice crypted Crypted Voice
(Analog) Voice Voice (Analog)
(Digital) (Digital)
50
Technical Approach
qStream Synchronization : SW
A/D
En
Or Crypted
Voice
D/A
Technical Approach
51
3. Stream Cipher System
n System Blocks (See Fig.2)
n Block (1) is codec interface or KMF identification interface
circuit, and
n Block (2) is circuit with CAI function.
n Block (3) is main controller to control whole system and
n Block (4) is a synchronization pattern generator which
generates encryption synchronization pattern.
n Block (5) and (12) are transmission and reception session
key buffer.
n Block (6) and (11) are transmission and reception session
key construction in order to distribute session key in secure.
n Block (7) and (13) are transmission and reception encryptio
n algorithm which have high security (i.e. SEED, AES).
n Block (8) and (14) are ZS (Zero-suppression), encryption/
decryption operation (XOR).
n Block (9) is data selector to divide among synchronization
pattern/session key/cipher text.
n Block (10) is synchronization pattern detector to detect patt
ern in transmission occurred
52
3. Stream Cipher System
q Encryption-SYNC Performances
Ø PD : Detection probability
Ø PM : Missing probability
Ø N : Detection window ,
Ø B : Channel bit error rates
Ø R (bps) : Transmission rates
Ø NT(0≤NT≤N) : Threshold
53
3. Stream Cipher System
q [Table] Sync. probability for variable NT(@BER = 10-1).
NT PF PD PM
0 2.938735877055719e-39 1.390084229174165e-06 9.999986099157708e-01
1 3.790969281401877e-37 2.116017137142591e-05 9.999788398286286e-01
2 2.426514213684907e-35 1.606491218512547e-04 9.998393508781488e-01
3 1.027479040902622e-33 8.115975682014417e-04 9.991884024317985e-01
4 3.237791337733303e-32 3.071835266561997e-03 9.969281647334380e-01
5 8.098686849208071e-31 9.300045916275050e-03 9.906999540837249e-01
6 1.674842950156203e-29 2.348652596439370e-02 9.765134740356063e-01
7 2.945347751630233e-28 5.095875762354045e-02 9.490412423764596e-01
8 4.496053253292625e-27 9.712736992623532e-02 9.028726300737647e-01
9 6.051629962835398e-26 1.655253152108301e-01 8.344746847891700e-01
10 7.271572314915841e-25 2.559625999178171e-01 7.440374000821830e-01
11 7.878396318751689e-24 3.637565370098624e-01 6.362434629901376e-01
12 7.760297741953771e-23 4.805333041263665e-01 5.194666958736335e-01
13 6.997607780111669e-22 5.963119811331635e-01 4.036880188668365e-01
14 5.810342711442406e-21 7.019829976111680e-01 2.980170023888320e-01
15 4.465076540551982e-20 7.912163018922382e-01 2.087836981077618e-01
16 3.189612506824416e-19 8.612396598832773e-01 1.387603401167227e-01
17 2.126183271330397e-18 9.124985894574300e-01 8.750141054256999e-02
18 1.327071906532612e-17 9.476204491915974e-01 5.237955080840262e-02
19 7.779171576740664e-17 9.702134587513881e-01 2.978654124861191e-02
20 4.294311477937454e-16 9.838947814335587e-01 1.610521856644131e-02
21 2.237862512500631e-15 9.917126802385244e-01 8.287319761475564e-03
22 1.103341505902957e-14 9.959375044101476e-01 4.062495589852388e-03
23 5.156943983868469e-14 9.981009409782960e-01 1.899059021703953e-03
24 2.289145482496759e-13 9.991526115496695e-01 8.473884503304996e-04
25 9.666701992393990e-13 9.996387170662638e-01 3.612829337361623e-04
26 3.889317585852534e-12 9.998526865920681e-01 1.473134079319482e-04
27 1.493042993527993e-11 9.999425009624112e-01 5.749903758878183e-05
28 5.475729948142875e-11 9.999784979923908e-01 2.150200760919763e-05
29 1.920913323991833e-10 9.999922899581286e-01 7.710041871389350e-06
30 6.452936410277733e-10 9.999973470123162e-01 2.652987683759989e-06
2013-05-09 107 Dongseo University,
Korea
54