Steganography Seminar

8/10/2019 Steganography Seminar

1/27

SteganographySTEGANOS() GRAPHEIN()


2/27

Introduction

Difference between cryptography and Steganography

We all know cryptography is about making information unreadablThen what is steganography be called ?

So what is it all about?

Technologies used.


3/27

History

Its been around for centuries

Slaves GreeksWax tablets

WWI and WWII

French resistance during WWIIInvisible ink

Washington and The iron lady .

Movies ,Microdots ,Printers

Modern dayDigital Steganography

Xerox and HP encoded printer serial numbers, as well as date andtime stamps in small yellow dots.


4/27

SteganographyBasic Framework


5/27

BASIC STEGANOGRAPHY MODEL

COVERIMAGE/AUDIO/VIDEO

(X)

MESSAGE (M)

STEGO-KEY (K)

F(X,M,K)

S

IMAGVI


6/27

GOOD

STEGANOGRAPHY

Robustness

CapacitySecurity


7/27

LSB Steganography

Basic concept = Substitution of cover bits with the message bits Some DIP concepts

Pixels

RGB Images = 8+8+8 =24bits/pixel

If we are using least 4 bits ,Does it mean we are using half theinformation of the cover image?

Concept of Perceptibility GoalMaximize capacity without compromising on security (here

the imperceptibility )

Solid background imagePoor choice of cover image

WHY??


8/27

Can you tell me the difference ??


9/27

Which one do you think wouldmake a good cover image ? Why?


10/27

You can replace 1 to 7 bits


11/27

Histogram Analysis

Histograms of mandrill and DalmatianCommon feature of color imageSmooth histograms


12/27


13/27

Mandrill Bit Planes


14/27

Steganalysis of LSB method

Visual Attack

If an embedding is detected through color abnormalities thesteganographic algorithm has been successfully attacked

An embedding is typically evidenced by a localized section ofnoise in the least significant plane.


15/27

STATISTICAL ATTACKS

Chi-Square Test

Histogram analysis of Pairs of Values (PoVs) that are swappedduring message embedding process.

The PoVs can be formed by pixel values, quantized DCTcoefficients that differ in the least significant bit

The distribution of occurrences of the two values from each pair whave a tendency to become equal after message embedding


16/27

HistogramsBefore and after embedding


17/27

Chi square Attack Algorithm

POV3 ALGORITHM

Let X (1281) and Y (1281) be two vectors such that xk =frequency(2k) and yk = frequency(2k + 1), 0 k 127. Initially, everyentry in X and Y is set to 0.

Then pov3 counts the gray values in the test image and increments thcorresponding entry in X or Y

The theoretically expected frequency15 of gray values 2k and 2k + 1 zk = (xk + yk) /2

Property: Sums of the frequencies of pixel values 2k and 2k +1 in acover image and in a stego image are the same, i.e., xk + yk in acover image equals xk +yk in its corresponding stego image


18/27

Whenever the combined frequency of 2k and 2k +1 is less than or equaindividual frequency counts of 2k and 2k +1 are set to 0 and the number

categories n is decremented by 1. The Chi-squared statistic, with n1 defreedom, is then calculated:

The expectation is that for a stego image, 2 n1is relatively small becaube near zi, by the hypothesis, and for a non-stego image, 2 n1 is relativbecause xi should be far from zi. The final step of the process is calculatinprobability of embedding, by integrating the density function with2 n-1upper limit


19/27

This probability of embedding is the probability of 2 n1

under the condition that xi = zi for all i in Equation 1. The density functio1p, converges to 1 as 2 n1 approaches infinity, so p approaches 0 a2 n1 approaches infinity. Therefore, for large 2 n1 , the probability oembedding is near 0. However, when 2 n1 is small relative to n 1, 1p is near zero and hence p is near 1. Thus for (relatively) small 2 n1 ,the probability of embedding is near 1.

In addition, Westfeld and Pfitzmann claimed that if less than 100% ofthe pixels contain embedded information, then the probability ofembedding will drop offsharply when a higher percentage of pixels istested.


20/27

Bit Plane Complexity Segmentation

In BPCS, the vessel image is divided into informative region andnoise-like region and the secret data is hidden in noise blocks ofvessel image without degrading image quality.

In LSB technique, data is hidden in LSB bits

In BPCS technique, data is hidden in MSB planes along with the LSBplanes provided secret data is hidden in complex region.

Randomization of the secret data by a compression operationmakes the embedded data more intangible

It is most secured technique and provides high security


21/27

Bit Plane SlicingThe Basic Concept


22/27

BPCS Algorithm1)Convert the carrier image(of any file format ) from PBC to Cand in png format. In BPCS-Steganography embedding oper

executed after the vessel image has been transformedfrom PBCto CGC.

2) Perform the histogram analysis and bit-plane analysis .

3) Perform size-estimation i.e. calculate the places where we the secret image. In size estimation we have to calculate the

where maximum color variations are observed. After doing thto store pixel value of secret image at that variation regions.


23/27

4)We segment each secret file to be embedded into a series of blocks havbytes of data each. These blocks are regarded as 8 8 image patterns. Wesuch blocks the secret blocks. We replace the complex 8 8 pixels block bmessage bits

Carrier is complex ..Ok.

Is that the only criterion??

What if the message data itself is not complex enough ??


24/27

During extraction, it will not exceed the threshold and hence wont beextracted !!

How we solve that ??

XOR with a checker board pattern

It conjugates the message data and hence guarantees messagecomplexity.


25/27

How to determine if the data obtained is conjugated

One solutionReserve one bit as indicator in each 8*8 matrix of pixels

Uses 1/64thcapacity

5)Convert the embedded dummy image from CGC back to PBC

6) De-steganography: De-steganography is exactly opposite ofsteganography. Here we will extract secret image from vessel image. In thway we will get the secret image form hiding it from the third person


26/27

Reference THESIS ON LEAST SIGNIFICANT BIT EMBEDDINGS: IMPLEMENTATION AND DETECTION ~ AARO

MILLERMAY 2012

PAIRS OF VALUES AND THE CHI-SQUARED ATTACK CHRISTY A. STANLEY DEPARTMENT OFMATHEMATICS, IOWA STATE UNIVERSITY MAY 1, 2005

WESTFIELD, ANDREAS AND ANDREAS PfiTZMANN. ATTACKS ON STEGANOGRAPHIC SYSTEM

BREAKING THE STEGANOGRAPHIC UTILITIES EZSTEGO, JSTEG, STEGANOS, AND S-TOOLSAN

LESSONS LEARNED. 3RD INTERNATIONAL WORKSHOP ON INFORMATION HIDING (2000).

STEGANOGRAPHY USING BPCS TECHNOLOGY ~RESEARCH INVENTY: INTERNATIONAL JOUR

ENGINEERING AND SCIENCE VOL.3, ISSUE 2 (MAY 2013), PP 08-16

PRINCIPLE AND APPLICATIONS OF BPCS-STEGANOGRAPHY EIJI KAWAGUCHI* AND RICHA

EASON KYUSHU INSTITUTE OF TECHNOLOGY, KITAKYUSHU, JAPAN ** UNIVERSITY OF MAINE,

MAINE 04469-5708

WIKIPEDIA ,ONLINE LECTURES AND ARTICLES ON STEGANOGRAPHY AND STEGANALYSIS


27/27

THANK YOU ALL !

Steganography Seminar

Documents