Fuzzy Inference System (FIS)

Download Report

Transcript Fuzzy Inference System (FIS)

Image Steganography Using Fuzzy Domain
Transformation and Pixel Classification
Aleem Khalid Alvi
Robin Dawes
School of Computing
Queen’s University, Kingston, Canada
Contents
Information Hiding Methods
Proposed Technique
Fuzzy Image Representation and Domain Transformation
Gaussian Membership Function (GMF)
Fuzzy Inference System (FIS)
Methodology
Using Image Processing for Fuzzy Pixel Classification
Analysis and Results
Conclusion
2
Information Hiding Methods
Attributes for classification of information hiding methods
 Cover objects
 Secret objects
 Hiding techniques
 Current technologies
Steganography system characteristics
 Robustness
 Security
 Un-detectability
 Imperceptibility (invisibility or perceptual transparency)
 High capacity
3
Proposed Technique
Proposed technique is the combination of
 Domain transformation
 Data conversion
 Substitution
 Image properties
It is kind of private-key steganography technique
4
Fuzzy Image Representation and Domain Transformation
An image representation in spatial and fuzzy
M
N
I s   I mn
m 1 n 1
M
N
 f    x ( I mn )
m 1 n 1
5
Gaussian Membership Function (GMF)
We use GMF for image
transformation from the
spatial domain into the
fuzzy domain
The specific image transformation function with fuzzifier
 ( I max  I mn ) 2
 mn  e
2 fh2
Where fh = fuzzifier, Imax = maximum pixel value of an image,
Imn = any gray level pixel value of an image I
6
Fuzzy Inference System (FIS)
We use Mamdani fuzzy interference system (FIS)
 Using the fuzzy inference process,
 A given input (a crisp input) maps to an output (a crisp output) using
fuzzy logic methods.
 The fuzzy inference process requires membership functions, logical
operations, and If-Then rules.
Implementation steps
 Fuzzify inputs
 Apply fuzzy operator
 Apply implication method
 Aggregate all outputs
 Defuzzify
7
Methodology
The step-by-step
methodological
information for
embedding process
on the sending end of
the steganography
System.
8
Using Image Processing for Fuzzy Pixel Classification
Use fuzzy based If-Then rules to apply fuzzy classification
Select the appropriate cover pixel for embedding secret data
Produce less disturbance and distortion in the embedded
cover image with respect to Human Visual System (HVS)
Use texture and silhouette (edge) properties of an image
9
Analysis and Results
Using Lena (Cover) and
Tomahawk Missile (Secret) Images
Lena.jpg available capacity
= 145,313 pixels
Secret data uses 17.62% of the
available capacity
HVS testing shows that original
and stego images have significant
difference and visible as light shades
Statistical testing shows differences
Cover histograms looks similar
10
Analysis and Results
Using Baboon (Cover) and
Tomahawk Missile (Secret) Images
Baboon.jpg has available capacity
= 138,518 pixels
Uses 18.48% available capacity
HVS testing shows that no
difference in visibility
Statistical testing shows the
difference between their statistical
values
Cover histograms looks similar
11
contd..
Conclusion
Proposed steganography algorithm based on fuzzy inference
system
Fuzzy inference system uses fuzzy transformation and pixel
classification techniques
The fuzzy pixel classification uses the image processing
techniques by exploiting texture and silhouette properties
The exploitation of the image processing techniques with
fuzzy logic increase imperceptibility in stego image
significantly
12
Thank You!
13