1. Using Particle Swarm Optimization to Embed Important Information in an Image Compression File
Chair Professor Chin-Chen Chang (張真誠)
National Tsing Hua University
National Chung Cheng University
Feng Chia University

2. Introduction Information Hiding Hiding system Stego image Cover image
Secret message

3. Introduction (Cont.) Cover Carriers Image Compression code Video Sound
Text

4. Kim et al.’s Method ： 9 1 2 3 4 5 6 7 8 …

5. Kim et al.’s Method (Embedding)5 8 3 4 7 6 1 2 ： 9 1 2 3 4 5 6 7 8
Cover Image
Cover Image 6 9 7 3 8 1 2 5 4 …
Stego Image
Stego Image

6. Kim et al.’s Method (Extracting)6 9 7 3 8 1 2 5 4 ： 9 1 2 3 4 5 6 7 8
Stego Image
Stego Image …

7. Zhang and Wang’s Method (Embedding)
Extracting function: 8 7 9 4 79 54 55 11 20 21 12 24 10
Secret data: … p2
255 1 2 3 4 1 2 3 4 1 : : : : : : : : : : : : :
10002
1 35 … 11 2 3 4 1 2 3 4 1 2 3 2 … 10 1 2 3 4 1 2 3 4 1
Cover image … 9 3 4 1 2 3 4 1 2 3 4 3 … 8 1 2 3 4 1 2 3 4 1 2 1 … 7 4 1 2 3 4 1 2 3 4 4 … 6 2 3 4 1 2 3 4 1 2 3 2 … 5 1 2 3 4 1 2 3 4 1 7 10 4 … 4 3 4 1 2 3 4 1 2 3 4 3 … 3 1 2 3 4 1 2 3 4 1 2 1 … 2 4 1 2 3 4 1 2 3 4 4 … 1 2 3 4 1 2 3 4 1 2 3 2 … 1 2 3 4 1 2 3 4 1
Stego image 1 2 3 4 5 6 7 8 9 10 11 …
255 p1
Magic Matrix

8. Zhang and Wang’s Method (Extracting)p2 7 10 4
255 1 2 3 4 1 2 3 4 1 : : : : : : : : : : : : : … 11 2 3 4 1 2 3 4 1 2 3 2 … 10 1 2 3 4 1 2 3 4 1 … 9 3 4 1 2 3 4 1 2 3 4 3 … 8 1 2 3 4 1 2 3 4 1 2 1
Stego image … 7 4 1 2 3 4 1 2 3 4 4 … 6 2 3 4 1 2 3 4 1 2 3 2 … 5 1 2 3 4 1 2 3 4 1 … 4 3 4 1 2 3 4 1 2 3 4 3 … 3 1 2 3 4 1 2 3 4 1 2 1 … 2 4 1 2 3 4 1 2 3 4 4
1 35 … 1 2 3 4 1 2 3 4 1 2 3 2 … 1 2 3 4 1 2 3 4 1 p1 1 2 3 4 5 6 7 8 9 10 11 …
255
Extracted secret data: 10002
Magic Matrix

10. Pixels in the Image Block
248 76 49 62 57 24 96
118
125
144 56 41 82 97
211
114

11. Generate Compression Code
248 76 49 62 57 24 96
118
125
144 56 41 82 97
211
114 1
AVERAGE: 100 H L
160 64

12. LSB Data Hiding SECRET 010 100 H L 160 64 H’ L’ 162 68 H L 10100 000 H’ L’
LSB Data Hiding

13. Recovery of the Image Block
160 64
162 68 H L
160 64 H’ L’
162 68

14. LSB Data Hiding SECRET 010 100 SECRET’ 001 010 H L 160 64 H’ L’ 161 66 H’ L’
LSB Data Hiding

15. Recovery of the Image Block
160 64
161 66 H L
160 64 H’ L’
161 66

16. Substitution Tables SECRET 010 100 2 4 SECRET ’ 100 110 4 6 Table 1 11 2 3 4 5 6 7
Table 2 1 2 3 4 5 6 7
Table 3 2 1 3 4 5 6 7
Table 40320 7 1 6 2 5 3 4

17. Example (1/3)
BTC
BM1
H1=100 L1= 37
BM2
H2=155 L2= 72
BM3
H3=212 L3= 86
BM4
H4=177 L4= 63
BM5
H5=151 L5 = 45
BM6
H6=132 L6 = 54
BM7
H7=201 L7= 73
BM8
H8=182 L8= 84
BM9
H9=106 L9= 42
BM10
H10=231 L10= 97
BM11
H11=214 L11= 98
BM12
H12=171 L12= 76
BM13
H13=145 L13= 65
BM14
H14=189 L14= 39
BM15
H15=179 L15= 74
BM16
H16=167 L16= 59
Compression codes CC0
S = =

18. Example (2/3)
By Table 1 → S =
BM1
H1=98 L1= 36
BM2
H2=155 L2= 77
BM3
H3=215 L3= 84
BM4
H4=177 L4= 56
BM5
H5=146 L5 = 44
BM6
H6=133 L6 = 55
BM7
H7=204 L7= 74
BM8
H8=180 L8= 83
BM9
H9=110 L9= 47
BM10
H10=228 L10=103
BM11
H11=214 L11= 97
BM12
H12=169 L12= 76
BM13
H13=147 L13= 69
BM14
H14=185 L14= 38
BM15
H15=176 L15= 74
BM16
H16=164 L16= 61
CC1 =
By Table 3 → S =
BM1
H1=100 L1= 38
BM2
H2=157 L2= 79
BM3
H3=209 L3= 86
BM4
H4=179 L4= 58
BM5
H5=148 L5 = 46
BM6
H6=135 L6 = 49
BM7
H7=206 L7= 76
BM8
H8=182 L8= 85
BM9
H9=104 L9= 41
BM10
H10=230 L10= 97
BM11
H11=214 L11= 99
BM12
H12=171 L12= 78
BM13
H13=149 L13= 71
BM14
H14=187 L14= 32
BM15
H15=178 L15= 76
BM16
H16=166 L16= 63
CC3 =

19. Example (3/3) Diff01 = 1193 Diff03 = 792 So, Table 3 is better!
Compute the difference between Recovery(CC0) and Recovery(CC1)
Diff01 = 1193
Compute the difference between Recovery(CC0) and Recovery(CC3)
Diff03 = 792
So, Table 3 is better!
Our job is to find the best table among all tables

20. Flowchart of PSO Algorithm
Initial particles If No
Yes
Update a particle If No
Yes
Compute fitness value No
Termination
Yes
End

21. Initial particles . . . Initial particles2 1 3 4 6 7 5 3 7 6 5 2 1 4 6 3 2 4 1 5 7
Global best particle:
Local best particle:
Our goal is to find the best particle that has the best fitness value!

22. Update a particle
Update a particle 2 1 3 4 6 7 5 7 1 3 4 6 2 5

23. Compute fitness value Fitness function Peak signal-to-noise ratio
Mean square error

24. Compute fitness value Fitness function LSB Data Hiding Stego image SI2 1 3 4 6 7 5
Table 2 1 3 4 5 6 7
LSB Data Hiding
Stego image SI

25. Particle
Fitness
value
25.4
26.8
27.6
28.5
29.2
28.6
27.9
27.8
Particle
New
fitness
Value
26.5
27.6
25.8
27.9
28.6
30.1

26. Experimental Results Secret message: LSB PSO ELSB Image PSNR EC Time
Airplane
29.96
0.375
31.04
309.38
31.06
Boat
29.62
30.68
328.45
30.71
Goldhill
30.65
31.73
294.27
31.74
Barbara
28.22
29.27
305.08
29.29
lena
30.92
32.01
310.96
32.02
peppers
30.91
31.98
313.04
31.99
Average
30.05
31.12
310.20
31.14
in matlab language
Secret message:

27. Conclusions
A novel method for embedding secrets in compression code based on PSO is presented.
The obtained substitution table is near optimal.
Higher embedding capacity always result in lower stego image quality.

28. Appendix