## Monday, 13 January 2020

### MATLAB code of DWT based Audio Steganography (Hiding Text into Audio)

Stegangraphy plays most important role in area of information security. Most of the researches are working on above topic. In this information era we need to protect the data by using various steganographic methods. In previous few post we seen

Now in this post we are focusing our approach on audio steganography using DWT and correlation.

The algorithm for same is as below
Cover media:
Given cover audio is .wav file
Let A is an original audio
Secrete Message:
Let  x be the secrete message in text. This message is 1st converted into binary 1D array
And represented as

Embedding Algorithm:
1st secrete text message is converted into binary string.

Step 1: Transformation of Cover Audio
Cover audio is transformed using DWT it will gives us two subbands
[ca,cd]=dwt(cover,'wname')
ca subband is low frequency band
cd contains diagonal details.
In this step, insertion of secret message onto cover object is carried out. Additional components rather than usual steganographic objects used here is pseudo-random number. Pseudo-random sequences typically exhibit statistical randomness while being generated by an entirely deterministic causal process generator. A pseudo-random number generator is a program that on input a seed, generates a seemingly random sequence of numbers.
Input: An 1 × n carrier audio and a secret message/image.
Output: An 1 × n stego-audio.

Algorithm: Steps-
1. Read the cover audio (Ic)

2. Calculate the size of Ic
3. Read the secret message (Im)
4. Prepare Im as message vector
5. Decompose the Ic by using Haar wavelet transform
6. Generate pseudo-random number (Pn)
7. Modify detailed coefficients like horizontal coefficients of wavelet decomposition by adding Pn when message bit = 0. And keep as it is when there is one.
8. Apply inverse DWT
9. Prepare stego audio.

Extraction Algorithm:

In this step extraction of secret message is carried out. Additionally correlation theory is being used. Correlation is the degree to which two or more quantities are linearly associated. The correlation between two same size matrices can be calculated by:
Input: An 1 × n carrier audio and an 1 × n stego-audio.
Output: a secret message

Algorithm:
Steps-
1. Read the cover audio (Ic)
2. Read the stego audio (Is)
3. Decompose the Ic and Is by using Haar wavelet transform
4. Generate message vector of all ones
5. Find the correlation between the original and modified coefficients
6. Turn the message vector bit to 0 if the correlation value is greater than mean correlation value
7. Prepare message vector to display secret image.

Obove algorithm is tested for various image quality parameters like Peak signal to noise ratio, Correlation, Mean Square Error and Total time required for algorithm execution.

MATLAB Implementation:

GUI Implementation
GUI Implementation

You-Tube Video