Docstoc

A Java Steganography Tool Hiding Behind Corners Using Edges in Images for Better Steganography

Document Sample
A Java Steganography Tool Hiding Behind Corners Using Edges in Images for Better Steganography Powered By Docstoc
					Hiding Behind Corners:
     Using Edges in Images for Better Steganography
                                    Kathryn Hempstalk
                Steganography
• Greek for “covered writing”
• Used by the Greeks to hide messages
   – In wax tablets
   – On the heads of slaves
• Some forms used in the last century are:
   – Invisible Ink
   – Microdots



                                             2
            Digital Techniques
Message: 010
Cover: 00001111 10101100 01000110
          byte

Stego-Message: 00001110 10101101 01000110

Similar to Cryptography and Watermarking
• Cryptography scrambles information
• Watermarking is more robust and not necessarily
  hidden
                                                    3
                       Used for…
•   Security purposes
•   Hiding keywords within files for better searching
•   Inline sound/subtitles
•   Information compression

• Great for images – as they naturally have a large
  amount of redundant information.
• Can hide as much as the image itself:
    capacity = (image.height() * image.width() * 3 * numLSB) / 8
                                                              4
     Basic Image Steganography

                            0   1   2   3   4   5   6    7       8
• Blind hiding – just
                        0   1   2   3   4   5   6   7    8       9
  starts at (0,0) and
                        1   10 11 12 .      .   .   ->
  writes until it’s
                        2
  finished              3
• Easy to break         4
                        5
                        6
                        7
                        8



                                                             5
An example…




              6
Where it put it…




                   7
         More Advanced Hiding

                               0   1    2   3   4    5   6   7       8
• Hide and Seek – just
                           0                             6
  randomly place it on
                           1       1        5
  the image
                           2
• More secure and hard     3   8                9                    10
  to break, but leaves a   4   7
  lot of noise on the      5                                 4
  image                    6            2       12
                           7
• Random seed used as
                           8       11                    3
  key to find data

                                                                 8
Hide and Seek example




                        9
Where it put it…




                   10
         Leaves “marks” behind




• A normal image has pixels where a single pixels
  neighbours are very similar in colour
• Laplace filter – measures noise by measuring how similar a
  pixel is to it’s neighbours
• LP(x,y) = 4(p(x,y)) –
                (p(x,y-1) + p(x,y+1) + p(x-1,y) + p(x+1,y))
                                                          11
 Using a filter for steganography
• Would be handy to be able to use the Laplace
  filter to help pick the best place to hide
• Problem: this requires the original image because
  the image changes after hiding
• Solution: keep some data the same
• Just use the lower order bits for steganography –
  higher order bits are used to obtain filter values
• Guarantees after hiding we can still find the best
  pixels again
                                                       12
         Filtering the Image First

                                  0   1    2   3   4   5   6   7        8
• Filter first and just run
                              0
  in the order of pixels in
                              1       1    2   3   4   5   6   7
  the filter
                              2       8                        9
• Highest filter values       3       10                       11
  are hidden in first         4       12                       13

• Still easy to break (no     5       14                       15
                              6       16 17 18 19 20 21 22
  key needed to find
                              7
  data)
                              8



                                                                   13
Filter first example




                       14
     Where it put it…




Laplace             Sobel   15
         Let’s play Battleships!

                                0    1   2    3   4   5    6   7        8
• The top 10% of values
                            0            14
  are the “ships”
                            1                     5   7    8   9        10
• Randomly choose pixel     2   13                6                     11
  until a “ship” is “hit”   3
• Then cluster the shots    4   2             4

  around the ship           5                         12
                            6                 16 15
• After a while, move
                            7            3
  away                      8                                           1



                                                                   16
Battleships example




                      17
     Where it put it…




Laplace             Sobel   18
                How good is it?
• Don’t want to look at the image to see
   – Too many images
   – Pixels are very small – can make it hard to tell
• Use a steganalysis method
   – Estimates the embedding rate in the LSB plane of the
     image
   – Very accurate results on current steganography
     methods (like BlindHide and HideSeek)
   – Can only apply to one image at time

                                                            19
  Machine Learning with WEKA
• Identify the steganographic pictures in an image set using
  support vector machines
• Use the RS Analysis, Sample Pairs Analysis and Laplace
  Values as attributes for prediction




                                                               20
    A Java Steganography Tool
• Can hide/retrieve using all the techniques shown
  here
• Can simulate the hiding on images
• Can benchmark using standard statistical
  techniques (mean squared error, average absolute
  difference…)
• Can perform RS Analysis, Sample Pairs Analysis
  and calculate Laplace Values.

                                                     21
Can you pick the stego-image?




                                22
http://diit.sourceforge.net
                              23

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:45
posted:9/6/2011
language:English
pages:23