HIGH professor, Student Electronics and communication department, Matrusri


Vijetha Kura, Buchhibabu Rachakonda

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now

Assistant professor,

Electronics and
communication department, Matrusri engineering college, Hyderabad, India

Electronics and
communication engineering , Matrusri engineering college ,Hyderabad ,India


paper will present a highly secured, very large capacity audio watermarking
system in which watermark/data i.e. an image can be encrypted and hidden by
modifying the magnitude values in FFT spectrum. The main idea is to modify
magnitude of FFT samples with respect to Fibonacci series .Here an image will
be encrypted and is embedded into audio which improves the security of
watermarking drastically. XOR sum of image bits and PN sequence is used as
embedding bit stream and is chosen as it is fastest and simple in computation
when compared to other encryption techniques. This technique mathematically proves
that maximum changes in FFT samples is less than 61% and the average error rate
considering a single sample is 25% .This technique is not only robust and
transparent but also highly secured. The additional feature is its ability to
handle large capacity i.e from experimentally its proved to handle 700bps to
3kpbs efficiently, moreover this technique is blind, i.e. original signal is
not required near receiver.


Index Terms:  audio watermarking , FFT, image encryption,
XOR sum,



In this
progressive era in which everything changes faster than the prick of the eye,
inventions followed by exploitation of its weakness is a casual scenario. One
of the important affected field is digital audio. As distribution of audio has
became very easy, which paved way for illegal distributions by which huge
intellectual scholars, authors suffered heavy losses due to copyright
infringements by illegitimate methods followed by criminals.

watermarking is a simple process in which a watermark which can be audio/Image
can be embedded into the host signal and can be later extracted and used for
multiple purposes.

watermarking has four different significant properties.

It is defined by the quality of the embedded signal i.e after adding watermark
in terms of objective and subjective measures.

The basic theme of the security is it should broadcast any clue from the
embedded signal. Security of a watermark defines how well it is ready to face
different attacks. The stronger the encryption the stronger is the security.

Robustness of and audio watermarking is defined by its ability to withstand
different types of attacks on embedded signal.

Here the payload is simply watermarking bits .It’s usually measured in bps i.e.
bits per second. The payload can be defined as the number of bits that can
embedded into host audio signal without losing significant imperceptibility of
the audio.

are different techniques available which includes D.C.T, M.D.C.T, Walsh
Hadamard-D.C.T, and D.W.T, in which imperceptibility and reduction of noise are
considered as main theme. This paper considers all properties i.e. imperceptibility,
security, robustness and payload and stands at the center of the tradeoff

can be embedded using different techniques.



can be embedded in frequency domain using different transforms i.e. FFT, DCT
,MDCT , Arnold DWT,DWT-DCT etc. out of which FFT is simple in complexity and
fastest out of them. One of the added advantage of FFT is its translation
invariant property.




2.1 Fibonacci series:

series origin takes place us to a scenario when Leonardo Fibonacci was resting
in a garden and watching rabbits, wondering how many rabbits would be born in
future if two rabbits mate and multiplication take place into their next

wonder number has distinguished qualities which has multiple applications which
include apple design logo, Benz logo .Our ear has a unique shape which can be
attributed to Fibonacci series.

this logos and different designs are designed using golden ratio i.e. 1.618, if
two numbers/Quantities are defined are golden ration if their ratio is equal to
the sum of the quantities to larger quantities.

a, b (a>b)

these two quantities are said to be in golden ratio if

(a+b)/a = a/b                                 


1: The typical maximum distortion that gets embedded in the FFT samples
(magnitude) using this algorithm is between the span of 0.38 to 0.61

if ‘l’ is converted to Fn+1

Max error rate = Max error / Fn+1




‘l’is converted to Fn

Max error rate =Max error/Fn


If we assume the general /typical value of Rn it is i.e golden ratio the max
error would be between 0.38-0.61. it indicates that the maximum error rate is

if we consider that fft values have equal probabilities then the average error
rate is 0.25 which indicates the average change per fft value is 25% only.

it has good imperceptibility.

2.2 Image encryption:

are different types of image encryption process available for different
purposes. Some of the famous techniques are chaotic image encryption, rubics
cube based image encryption ,steganography and many more out of which
encrypting with PN sequence is simpler and faster one. As our main concern is
to provide good security with fast computation encryption with PN sequence is

is the art of hiding information in a image, this can be done by varying the
gray values of the original image and then encrypting the information in it.
Now a days it is popular because of its robustness.

encryption can be of three types i.e. block permutation in which image is
divided into separate blocks and then permutated, in pixel permutation the
image pixels are permutated to create encrypted image.

block permutation image is divided into multiple parts and they are permutated
with proper techniques and they are inserted embedded into host signal

pixel permutation the image pixels are permutated with a secret key and them
embedded in the host signal .This method is widely used due to its robustness
and high fidelity.

bit permutation bits are permutated and then embedded into host signal, this
method is highly reliable due its complexity to decrypt compared to others.

 In this paper first the pixel gray value are
noted down and then converted into bits, Then XOR sum of grey values and
pseudo-random bit stream is generated.

simple XOR of watermark bits and PN sequence would produce same number of
watermark bits as original but XOR sum of watermark bits and PN sequence would
reduce drastically, the total number of watermark bits present after encryption
when compared to initial watermark bits available.

helps in increasing the payload or capacity of the audio watermarking without a
significant side effects.

generated bit stream is embedded into different frames of FFT coefficients that
are created after choosing required parameters (Frame size and bandwidth). The
FFT co-efficient are manipulated or changed w.r.t Fibonacci numbers and the bit
that is going to get embedded. This is main principle of embedding bits.


2.3 Tuning:

quality of watermarked audio is decided by few parameters i.e. Objective
degradation (ODG) , BER (Bit error rate) , payload (capacity) .,etc.

parameters can be varied to required values by altering two characteristics of
the algorithm i.e frequency bandwidth (Fl,Fh) and frame size (d).

– Lower frequency limit

– Higher frequency limit (Default value of fl=12 kHz and fh=16 kHz and d=5 is
considering Human auditory response).

default value can be as low as 10 kHz considering Human auditory response,
similarly 16 kHz is an average peak frequency in most of the audios.

by setting default parameters we should vary the characteristics as per our
requirements .If we look carefully we can observe that all the parameters are
interlinked to each other. This indicates the tradeoff triangle .Security, functionality,
imperceptibility are the three corners of the tradeoff triangle. This trade of
triangle is limited only to certain frequency bands and frame sizes, i.e.
varying them smartly we can overcome the inefficiency in tradeoff triangle.

Below figure describes about tuning and its
interrelation between different parameters.


                                                                             Fig 2.1
Tuning parameters




paper is extension to audio water marking using Fibonacci series , Here instead
of embedding simple bits we are embedding an image into the audio (Encrypted

is achieved by first encrypting the image with PN sequence generated by PN
generator (PNRG) i.e. XOR sum of image bits and pseudo random bit stream and
embedding them into frames that are obtained after applying FFT to original audio.
By using XOR sum the payload or capacity can be increased drastically.


3.1 Watermarking process:


3.1.1 Encryption:

First the frame size and frequency band length will be received securely to the

First convert the given audio into frequency domain signals, i.e apply FFT

Now select the co-efficients which falls between the selected frequency band.(
Fl, Fh)

The above step can be completed using different bandpass filters

After filtering, arrange all the co-efficients according to the frame size d.

Here 0 and 1 are removed from Fibonacci series in this process i.e.


k=1, 2, 3, 4, 5,… n.

Now add the watermark signal to the FFT co-efficient according to Fibonacci
numbers and bit that to be embedded.

7) Formulae:

 f ‘ = fib(k,i) , if k modulus 2=0 and wl=0

         fib (k+1,i) , if k modulus 2 =1 and


f’ =
fib(k+1,i) , if k modulus 2 =1 and wl=1

       fib (k,i) , if k modulus 2 =1 and wl=1

k represents kth Fibonacci number

Repeat the same process to all the FFT co-efficients in the frame

Repeat the same process to all the frames available.

Finally apply IFFT

this step encryption is completed


3..1.2 Extracting /Decrypting:

Apply FFT to watermarked signal as the operations are to performed in frequency

Divide the samples with given frame size d

Now change the FFT magnitude of given samples approximating to Fibonacci series
according to given formulae

Formula: D(i)=  0 , if k modulus 2 = 0

                                  1 , if k
modulus 2 =1


Now by polling method we can decide whether it is 1 or 0 , if the number of
samples found as zero out of half of the samples present in the frame , Then
its is considered as 0 else 1 .

there are 6 FFT co-efficients in a frame and 4 of the FFT Co-efficients when
decoded gives zero then the water bit embedded in the frame is ‘0’.

Now re-frame the encrypted image by using the extracted watermark bits

Now decrypt the image by using PRNG i.e XOR sum (same method) , The decryption
cannot be done without the seed of XOR

8) We
will get the decrypted image as output.



3.1 Encryption flowchart





4.1 Transparency,
Robustness, capacity, security:

to noise ration and ODG is used to measure imperceptibility, I.e greater the
SNR greater the imperceptibility similarly if ODG=0 ODG (Objective
degradation).it indicates that there is no degradation, if ODG=-4 then it
indicates very annoying distortion is present in the watermarked signal.

SDG (Subjective degradation) is five-point subjective grade i.e if S.D.G =1 it
is excellent and if S.D.G =1 it’s ridiculous

(Bit error rate):%BER = Number of error bits / Total number of bits

first layer of security in this algorithm is its frame size and frequency band length,
without knowing these two parameters it is merely impossible to decrypt the

second layer of security is the seed of PNRG that is required at both at sender
and receiver end without which decryption would be practically impossible.

embedding rate or capacity can be increased by either increasing the frequency
bandwidth (Fl,FH) or decreasing the frame size.

has little side effects as some of the security is sensitive to certain attacks
if frequency bandwidth is significantly increased.

 Below table shows the trade between capacity,
transparency and frame size which decide the quality of encryption.



4.1 Results of 5 Mono signals



7.2 Robustness test results of two selected files

 Below indicates the reason why we have chosen
only Fibonacci series in this encryption algorithm.



                           Table 7.3
Comparisons between different series that are generated with different k values
(K- ratio)


 Below table show comparison of different methods



                 Table 7.4 Comparisons between different


The above tables are constructed with reference to 27




this paper a high security, high capacity, robust, transparent, algorithm to
encrypt watermark into audio is presented. The image is encrypted by generating
bit stream of sum of XOR of pixel values and pseudo random sequence (generated
by PRNG) and this method is blind as we do not require original signal during decryption.
By using XOR sum for encryption the payload capacity is drastically increased.
The two deciding factors to change the above parameters are the frame size and
frequency bandwidth This paper provided proof that the maximum change of FFT
samples is less than 61% but average change of a single FFT sample is just
below 25%.Experimental proofs shows that this algorithm can handle capacity
ranging from 700bps to 3kbps and is robust against all common signal processing




1 M. Fallahpour and D. Megías,
“Robust high-capacity audio watermarking based on FFT amplitude modification,” IEICE
Trans. Inf. Syst., vol. E93-D, no. 01, pp. 87–93, Jan. 2010. 1282
8, AUGUST 2015

2 M. Fallahpour and D. Megías,
“Secure logarithmic audio watermarking scheme based on the human auditory
system,” Multimedia Syst., 2013, DOI: 10.1007/s00530-013-0325-1,

3 S. T. Chen, G. D. Wu, and H. N.
Huang, “Wavelet-domain audio watermarking scheme using optimisation-based
quantisation,” IET Signal Process., vol. 4, no. 6, pp. 720–727,

4 No, Really, “Rust,” Online.
Available: http://www.jamendo.com/en/album/7365 Jul. 17, 2014

5 T. Thiede, W. C. Treurniet, R.
Bitto, C. Schmidmer, T. Sporer, J. G.

Beerens, C. Colomes, M. Keyhl, G.
Stoll, K. Brandenburg, and B. Feiten, “PEAQ – The ITU standard for objective
measurement of perceived audio quality,” J. AES, vol. 48, no. 1/2, pp.
3–29, 2000.

6 S. T. Chen, H. N. Huang, C. J.
Chen, and G. D. Wu, “Energy-proportion based scheme for audio watermarking,” IET
Signal Process., vol. 4, no. 5, pp. 576–587, 2010.

7 G. Hua, J. Goh, and V. L. L.
Thing, “Time-spread echo-based audio watermarking with optimized
imperceptibility and robustness,” IEEE/ACM Trans. Audio, Speech, Lang.
Process., vol. 23, no. 2, pp. 227–239, Feb. 2015.

8 R. A. Dunlap, The golden
ratio and fibonacci numbers. Hackensack,

9 J. Katz and Y. Lindell,
“Introduction to modern cryptography: Principles and protocols,” in Chapman
& Hall/CRC Cryptography and Network Security Series. Boca Raton,
FL, USA: CRC, 2007.

10 H. J. Kim, “Audio watermarking
techniques,” in Proc. Pacific Rim Workshop Digital Steganogr., 2005, pp.

11 S. Xiang, H. J. Kim, and J.
Huang, “Audio watermarking robust againsttime-scale modification and MP3
compression,” Signal Process., vol. 88, no. 10, pp. 2372–2387, Oct.

12 M. Mansour and A. Tewfik,
“Data embedding in audio using time-scale modification,” IEEE Trans. Speech
Audio Process., vol. 13, no. 3, pp. 432–440, May 2005.

13 J. J. Garcia-Hernandez, M.
Nakano-Miyatake, and H. Perez-Meana, “Data hiding in audio signal using
rational dither modulation,” IEICE Electron. Express, vol. 5, no.
7, pp. 217–222, 2008.

14 M. Fallahpour and D. Megías,
“High capacity audio watermarking using FFT amplitude interpolation,” IEICE
Electron. Express, vol. 6, no. 14, pp. 1057–1063, 2009.

15 M. Fallahpour and D. Megías,
“High capacity method for real-time audio data hiding using the FFT transform,”
in Advances in Information Security and Its Application. Berlin,
Germany: Springer-Verlag,

2009, pp. 91–97.

 16 M. Fallahpour and D. Megías, “DWT–based
high capacity audio watermarking,” IEICE Trans. Fundam. Electron., Commun.
Comput. Sci., vol. E93-A, no. 01, pp. 331–335, Jan. 2010.

17 W. Li and X. Xue, “Content
based localized robust audio watermarking 8, no. 1, pp. 60–69, Feb. 2006.

18 M. Fallahpour and D. Megías,
“High capacity audio watermarking using the high frequency band of the wavelet
domain,” in Multimedia Tools and Applications. New York, NY, USA:
Springer, 2011, vol.

52, pp. 485–498.

 19 M. Fallahpour and D. Megías, “High
capacity robust audio watermarking scheme based on FFT and linear regression,” Int.
J. Innovative Comput., Inf. Control, vol. 8, no. 4, pp. 2477–2489,
Apr. 2012.

 20 N. K. Kalantari, M. A. Akhaee, M. Ahadi,
and H. Amindavar, “Robust multiplicative patchwork method for audio
watermarking,” IEEE Trans. Audio, Speech, Lang. Process., vol.
17, no. 6, pp. 1133–1141, Aug. 2009.

 21 X. Kang, R. Yang, and J. Huang, “Geometric
invariant audio watermarking based on an LCM feature,” IEEE Trans.
Multimedia, vol. 13, no. 2, pp. 181–190, Apr. 2011.

21 , G. E. Bergum, Ed. et al.,
Applications of Fibonacci Numbers. New York, NY, USA: Springer, 1991, vol.

 22 OPTICOM OPERA software site, Online.
Available: http://www.opticom. de/products/opera.html

23 Stirmark Benchmark for Audio,
Online. Available: http://wwwiti.cs. uni-  22 M. Unoki and D. Hamada, “Method of
digital-audio watermarking based on cochlear delay characteristics,” Int. J.
Innovat. Comput., Inf. Control, vol. 6, no. 3(B), pp. 1325–1346,
Mar. 2010.

24 K. Kondo and K. Nakagawa, “A
digital watermark for stereo audio signals using variable inter-channel delay
in high-frequency bands and its evaluation,” Int. J. Innovat. Comput., Inf.
Control, vol. 6, no. 3(B), pp. 1209–1220, Mar. 2010.

25 A. Nishimura, “Audio data
hiding that is robust with respect to aerial transmission and speech codecs,” Int.
J. Innovat. Comput., Inf. Control, vol. 6, no. 3(B), pp. 1389–1400, Mar.

26 Chen, Brian, and G. W.
Wornell, “Quantization index modulation: A class of provably good methods for
digital watermarking and information embedding,” IEEE Trans. Inf. Theory,
vol. 47, no. 4, pp. 1423 1443, May 2001.

 27 Mehdi Fallahpour and David megias “Audio
watermarking using fibonacci series”,
IEEE/ACM trascations on audio speech and language processing vol 23 ,no
8,August 2015.

28 X. Y. Wang and H. Zhao, “A
novel synchronization invariant audio watermarking scheme based on DWT and
DCT,” IEEE Trans. Signal Process., vol. 54, no. 12, pp.
4835–4840, Dec. 2006.



Vijetha Kura is currently
assistant professor in matrusri engineering college had completed M.tech and
has 8+ Years’ experience in teaching field , published 5 papers on different
topics and  the areas of research interest
are VLSI , Digital Communication , Digital encryption techniques , Data
processing techniques.


Buchhibabu Rachakonda is a student
pursuing graduation in Matrusri engineering college, Hyderabad, India.