 Open Access
 Total Downloads : 123
 Authors : John Britto, M. Mani Roja
 Paper ID : IJERTV6IS090174
 Volume & Issue : Volume 06, Issue 09 (September 2017)
 DOI : http://dx.doi.org/10.17577/IJERTV6IS090174
 Published (First Online): 03102017
 ISSN (Online) : 22780181
 Publisher Name : IJERT
 License: This work is licensed under a Creative Commons Attribution 4.0 International License
Audio Watermarking with Encryption
Water Marking
John Britto Student, SPIT Mumbai, India
M.Mani Roja Professor, TSEC Mumbai, India
Secrete Image
Audio Input
AbstractWatermarking is an effective means of hiding data, thereby protecting the data from unauthorized or unwanted viewing. Watermarks have been proposed previously as a way for protecting the data. This digital signature could discourage copyright violation, and may help determine the authenticity and ownership of an image. In this paper, audio watermarking is done with the help of Discrete Fourier Transform (DFT). To enhance the security of the watermark, pseudo noise (PN) based encryption is used.
KeywordsWater Marking; DFT; Encryption; PN sequence;
DFT
Encryption
IDFT
Figure 1: Watermarking Embedding
O/P
Image Retrieval
DFT
Audio Input
MSE

INTRODUCTION
The block diagram for watermark extraction is shown in fig 2.
Watermarking is the practice of imperceptibly altering a Work to embed a message about that Work [1]. Watermarking is becoming increasingly popular, especially for insertion of undetectable identifying marks, such as author or copyright information to the host signal. Watermarking may probably be best used in conjunction with another datahiding method such as steganography, cryptography etc. Such data hiding schemes, when coupled with watermarking, can be a part of an extensive layered security approach.
Audio watermarking is defined as secure communication of data related to the host audio signal, which includes embedding into, and extraction from, the host audio signal [2]. Watermarking, being ideally imperceptible, can be essentially used to mask the very existence of the secret message [3]. In this manner, watermarking is used to create a covert channel to transmit confidential information [4]. Digital audio watermarking involves the concealment of data within a discrete audio file.

PROPOSED BLOCK DIAGRAM
Audio watermarking will be implemented using the following steps

Read the secret data to be hidden.

Encrypt the secrete data using PN sequence

Read the audio signal in which data is to be hidden.

Take DFT of the audio signal

Hide the data in the audio at the High frequency DFT coefficients.

Take IDFT
The proposed block diagram for audio watermarking is shown in fig 1.
Audio Output
Decryption
Figure 2: Watermarking Extraction
The extraction of hidden data at the receiver will be implemented using the following steps

Read the received Audio

Find DFT

Extract the watermark from the high frequency DFT coefficients

Decrypt the watermark

Audio Input
MATLABÂ® audio support provides the ability to:

Read and write audio files in common formats such as WAV, AVI, FLAC, MP3, and MPEG4 AAC

Playback and record audio files using the PC sound card
In this research, an audio file with .wav extension is considered as cover data and sampled at a rate of 44.1 kHz with 16 bit analog to digital converter.


Discrete Fourier Transform [5,6]
The Fourier transform is simply a method of expressing a function in terms of the sum of its projections onto a set of basis functions. The DFT is a specific kind of discrete transform, used in Fourier analysis. It transforms one function into another, which is called the frequency domain representation, or simply the DFT, of the original function (which is often a function in the time domain). The DFT requires an input function that is discrete.
The DFT transforms N complex numbers into another sequence of complex numbers, which is defined by F(k) where k=1,1 N1
N 1
F (k) f (n)e
n0

j 2kn N
(1)



IMPLEMENTATION
The input, i.e., spectrum of the original audio signal from the song Maa Tujhe Salam has been considered as cover audio. The cameraman image from MATLAB directory is
The inverse Fourier Transform can be calculated using the
following equation.
considered as watermark. The cover audio undergoes Discrete Fourier Transform and got converted back into frequency
domain. The maximum energy components lie near the origin
F (n)
1 N 1 F (k)e N n0
j 2kn N
(2)
and low energy components lie at higher frequencies. Hence it has been decided to embed the watermark in the high frequency components of cover audio.

Water mark Selection
Audio Watermarking can be implemented in 3 ways:

Audio in Audio

Audio in Image

Image in Audio


Need for Encryption[7,8]
Encryption is the process of transforming data into scrambled unintelligible cipher text using a key. The role of encryption is to secure information when it is stored or in transit. However, it is relatively easier to crack single level encryption by brute force or correlation, as compared to a multilevel encryption scheme. Over the years, Data transmission has been made very simple, fast and accurate using the internet. However, one of the main problems associated with transmission of data over the internet is that it may pose a security threat, i.e., personal or confidential data can be stolen or hacked in many ways. Therefore, it becomes very important to take data security into consideration, as it is

Need for Normalization
Since the cover signal is the audio signal, the amplitude variations are within Â±1V. Watermark is a grey level image with 256 quantization levels where the pixels values lies within the range [0, 255]. Hence it is necessary to normalize the grey values within Â±1 V so that without affecting the magnitude spectrum, the stuffing of grey values can be implemented.

Z – Score (ZS) Normalization[11]
Zscore normalization calculates the normalized scores using arithmetic mean and standard deviation of the given data. This method transforms the score s to a distribution with mean 0 and a deviation of 1. Let mean (s) denote the arithmetic mean of score range and std (s) denote the standard deviation operators, respectively. Then the normalized score
(N) is calculated as
N s mean(s)
one of the essential factors that need attention during the process of data distribution.
std (s)
(3)
Cryptography gained prominence during the Second World War when the allied forces gained an upper hand after they were able to break the German cipher machine, Enigma [9]. These days, it is recognized as one of the major components for providing information security, controlling access to resources and financial transactions. The original data to be transmitted is called plain text which is readable by a person or computer. When it is encrypted, it is known as Cipher Text. The level of security of an encryption algorithm is given by the size of its key space [10]. The larger the key space, the more complicated the encryption algorithm is.
Cryptography keys are usually classified as symmetric and asymmetric algorithms. In symmetric key algorithms, the sender and receiver use the same keys for encryption as well as decryption. Symmetric key encryption is also known as secret key as both, the sender and receiver have to keep the key protected [7,10]. The level of security entirelydepends on how well the sender and the receiver keep the key protected. If the unauthorized person is able to get the key, he can easily decipher the encryption using it. This is the major limitation of the symmetric key algorithm.
In symmetric key encryption, a secure mechanism is required to deliver keys properly while asymmetric keys result in better key distribution. Symmetric key provides confidentiality but not authenticity as the secret key is shared [8]. Asymmetric keys on the other hand, provide both authentication and confidentiality [7, 8].
After normalization, the high frequency components of cover audio are replaced with normalized values of watermark pixel grey values. After completing the process, IDFT is taken for the combined data so that the signal comes back to time domain. This signal almost resembles the original signal as shown in fig.3.
For the extraction of the watermark, the DFT of the received audio has been taken and from the high frequency components, the watermark gets extracted as shown in fig.3

PN sequence based Encryption[13]
PN sequences look like random noise but they are not purely random in nature. When we multiply the PN sequence with an image, the resulting image resembles noise. However, during decryption, this noiselike signal can be used to exactly reconstruct the original image by multiplying it by the same pseudorandom sequence. Using this scheme, the initial seed state which is nothing but the key is only needed to generate exactly the same sequence of length. We have decided the following algorithm to generate the seed state for the PN sequence generator.

Decide the password.

Generate the seed state from the password
Since the seed state is known only to the authorized user, an intruder trying to access the database will not be able to decrypt the templates.
Figure 3. Watermark Extraction
The encryption scheme using PN sequence is done as given below [13]:

Let the watermark be image K

Generate the Pseudo image R using the seed state. The size of R image should be same as image K

Do the bit ExOR operations on both the images and find out the resultant image C as
C= bit XOR(R, K) (4)
The decryption scheme using PN sequence is done with the help of following steps [13]:

Generate the Pseudo image R using the seed state. The size of R image should be same as extracted image C

Do the bit ExOR operations on both the images R and C

The watermark is extracted as

W= bit XOR(R, C) (5)
Figure 4: Encrypted Watermark extraction


RESULT ANALYSIS
For all practical purposes, it is preferable to have a quantitative measurement to provide an objective judgment of the extracting fidelity. This is done by calculating the following parameters in each case. Results, thus obtained, have been tabulated. For performance evaluation, an audio clipping with one minute duration is considered as cover data.

Mean Square Error
Mean Square Error (MSE) [14], first introduced by C. F. Gauss, serves as an important parameter in gauging the performance of the watermarking system. MSE is essentially a signal fidelity measure [15, 16]. The goal of a signal fidelity measure is to compare two signals by providing a quantitative score that describes the degree of similarity/fidelity. Suppose that
x = {xi  i = 1, 2, N} and
y = {yi i = 1, 2, N} are two finitelengths, discrete signals, The MSE between the signals are given by the following formula:
The watermarked image is encrypted using PN sequence and the resultant encrypted image is transmitted to the receiver along with cover audio. At the receiver side, first the encrypted watermark is extracted from the cover audio and then
MSE
1 ( X
N
i
N i1
Y )2
(6)
i
decryption using PN sequence is done on the extracted image as shown in fig.4
where, N is the number of signal samples, xi is the value of the
ith sample in x. yi is the value of the ith sample in y.

Entropy
The entropy H of a discrete random variable X is given as
H (x) E(I (x))
(7)
Here E is the expected value, and I is the information content of X. If p denotes the probability mass function of X then the entropy can be explicitly written as:
n n
H (x) p(xi )I (xi ) p(xi ) log2 p(xi )
i1
i1
(8)

Moment
In image processing, computer vision and related fields, an image moment is a certain particular weighted average (moment) of the image pixels intensities, or a function of such moments, usually chosen to have some attractive property or interpretation. For a gray scale image with pixel intensities I(x, y), raw image moments are calculated by:
REFERENCES

I. J. Cox, M. L. Miller, J. A. Bloom, J. Fridrich, and T. Kalker, Digital Watermarking and Steganography, 2nd Edition, Morgan Kaufmann, 2008, p. 31

N. Cvejic, Academic Dissertation, Faculty of Technology, University of Oulu, Algorithms for audio watermarking and steganography, p. 5.

S. Shiyamala and Dr. V. Rajamani, A Novel Area Efficient Folded
Modified Convolutional Interleaving Architecture for MAP Decoder, International Journal of Computer Applications (IJCA), Vol. 9, No. 9, p.
M ij
xi y j I (x, y)
x y
(9)
1, November 2010.

Shahreza S.S. and Shalmani M.T.M., Adaptive wavelet domain audio steganography with high capacity and low error rate, Proceedings of the
Table I. Performance Evaluation (1minute audio)
Parameter
Simple Watermark
Encrypted watermark
Entropy
7
7
Moment
1.17 109
1.17 109
MSE
0
0
Embedding Time
1.16s
1.49s
Extraction Time
0.18s
0.77s
MSE in Audio
2.2 104
5.11104


CONCLUSION
In this paper, an audio watermarking scheme based on Discrete Fourier Transform has been implemented. To enhance the security of the system, A PN sequence based encryption has also been considered. The proposed algorithms were implemented using MATLAB 2010a on an Intel core i3 based platform. For a cover audio of size one minute, the average time taken for extraction simple watermarking with an image of size 256 Ã— 256 was 0.18 seconds and 0.77 seconds using the PN sequence scheme. In terms of performance, the calculated values of MSE, entropy and moment are found to be the same for both the schemes. Since the MSE between original audio and watermarked audio was very high, the clarity of the audio at the receiver decreased. As a future scope, this MSE can be reduced with the help of some other watermarking techniques.
IEEE International Conference on Information and Emerging Technologies, (ICIET07), pp. 17291732, 2007.

Rafel C.Gonzalez and Richard Woods, Digital Image Processing, 9th edition, Printice Hall, NewYork, September, 2009.

Anil Jain, Fundamentals of Image Processing, 4th edition, Pearson publications, 2002.

Aman Chadha, Sandeep Gangundi, Rishabh Goel, Hiren Dave, M. Mani Roja, Audio Watermarking with Error Correction, (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 2, No. 9, 2011

William Stallings, Cryptography and Network Security, Pearson Education, 5th edition, 2011.

D. Kahn, Cryptography Goes Public, IEEE Communications Magazine, Vol. 18, 1928, 1980.

G. Wite, AllinOne Security+ Certification Exam Guide, McGraw
Hill, 2003

Anil Jain, Karthik Nandakumar and Arun Ross, Score normalization in Multimodal Biometric Systems, Pattern Recognition, vol.38, no.12, pp.2270 2285, 2005.

R. Mutagi, "Pseudo Noise Sequences for Engineers," Electronics & Communication Engineering Journal, vol.8, no.2, pp.7987, April, 1996.

L. Trevisan, "CS276 Lecture 5: Encryption Using Pseudorandom Functions", (in theory), [online] 2009, https://lucatrevisan. wordpress.com/2009/02/13/cs276lecture5encryptionusing pseudorandomfunctions/ (Accessed: 8 February 2015).

Z. Wang and A. C. Bovik, Mean squared error: love it or leave it? – A new look at signal fidelity measures, IEEE Signal Processing Magazine, Vol. 26, No. 1, pp. 98117, January 2009.

G. Casella and E.L. Lehmann, Theory of Point Estimation. New York: SpringerVerlag, 1999.

T. N. Pappas, R. J. Safranek and J. Chen, Perceptual criteria for image quality evaluation, in Handbook of Image and Video Processing, 2nd ed.,