JRMAT



XXXXXX (TITLE)

(AUTHORS NAMES) XXXXXX #1, XXXXX *2

XXXXXXXXXX(DESIGNATIONS)

XXXXXXX(COLLEGE NAME WITH MAIL IDS)

ABSTRACT--Steganography is art of hiding information into a cover object that can be an image, video or audio. Speech steganography is a process of hiding message data into a cover speech without degrading the quality of cover speech. In this, a novel user interface for spread spectrum representation-based speech steganography using translation invariant wavelet transform (TIWT). This is an extension for Fast fourier transform (FFT) and discrete wavelet transform (DWT) based steganography.Simulation results proved that the proposed algorithm is superior to the conventional algorithms. Also performed good enough simulations with low bit error rate and excellent imperceptibility.

Keywords---Digital Steganography, Spread Spectrum, Speech steganography, FFT, wavelet analysis, discrete wavelet transform and translation invariant transform.

I. INTRODUCTION

The principle target of digital steganography is to shroud a mystery message inside a digital spread protest, that means unauthorized people or others can't recognize the concealed message nearness [1]. Present day steganography utilizes the opportunity of concealing information into digital multimedia documents furthermore at the system parcel level. We require couple of components to conceal the information into a spread media as followed by.

• The spread media(C) to hold the information which is to be covered up

• The mystery message (M) i.e., any sort of information.

• The stego function (Fe) and its opposite (Fe-1)

• An discretionary stego-key (K) or watchword might be utilized to cover up and unhide the message [3].

Digital steganography is known as to shroud information on the front of digital items [3]. Spread protests that are utilized as a part of digital steganography can fluctuate, for instance in the image chronicle. Steganography algorithms in the image file have been generally created. Meanwhile, steganography algorithms in audio chronicle are moderately few. This paper talks about the utilization of digital steganography on audio documents utilizing the technique for Direct-grouping Spread Spectrum [1]. what's more, Discrete Wavelet Transform techniques. Steganography in the audio document is not as simple as in the image file. Unlike the documents of crude images, crude sound records are typically bigger. In correlation, the crude image record sort and resolution of 1280x800 24bit shading (standard resolution of desktop screen) has a size of around 3 MB of information. While the crude audio documents with 44.1 kHz inspecting frequency, 16 bit stereo channels with 4 minutes length (the standard term of tune) has a size of around 40 MB of information. The distinction is entirely extensive, bringing about the execution of steganography in audio information turns out to be more troublesome [4].

[pic]

Fig.1 block diagram of digital steganography

Fig1. Demonstrates that the block outline of digital steganography strategy which incorporates the spread item, secure key, mystery message as the information. At that point, after we will get a stego audio by applying any algorithm. Presently, the message will be separated and the audio additionally reproduced to its unique arrangement. Also, the utilization of crude audio documents (WAV) is less successive than the crude image records (BMP), on the grounds that the size in audio records is too substantial. In this manner, we need such a plan, to the point that empowers us to save the shrouded messages [5], regardless of the fact that the audio documents are compacted. In the following segment, the creator will clarify some fundamental speculations that should be known ahead of time. In our proposed algorithm we are utilizing Discrete Wavelet transform to change over the audio information space into frequency area of four sub bands (LL, LH.HL and HH). In this four sub bands LL alluded as Approximation Coefficients whereas staying three sub bands alluded as point of interest coefficients [5].

II. EXISTING WORK

In this segment, we examined Fast Fourier Transform (FFT) based steganography plan proposed in [6]. Also, clarified the reasons why the creator has utilized FFT rather than Fourier and short time Fourier transforms.

A. Fourier Transform

FT is a methodology which separates the sign into various frequencies of sinusoids and it is characterized as a scientific methodology for transforming the sign from time space to frequency area. FT has a downside that it will work out for just stationary signs, which won't change with the time frame. Since, the FT connected for the whole flag however not portions of a sign, on the off chance that we consider non-stationary sign the sign will fluctuate with the time frame, which couldn't be transformed by FT. also, one more downside that we have with the FT is we can't say that at what time the specific occasion will has happened.

B. Short-Time Fourier analysis

To adjust the inadequacy in FT, Dennis Gabor in 1946 presented another technique called windowing, which can be connected to the sign to investigate a little area of a sign. This adjustment has been called as the Short-Time Fourier Transform (STFT), in which the sign will be mapped into time and frequency information. In STFT, the window is altered. In this way, we this window won't change with the time of the sign i.e., for both tight resolution and wide resolution. Also, we can't anticipate the frequency content at every time interim segment.

C. FFT Algorithm

This section explains a FFT domain Speech steganography using spread spectrum. Here, FFT is used to transform the audio cover object into the time domain to frequency domain.Then the signal information will be added to the cover signal by using the spread spectrum. Input message which is to be embedded into the cover object will be converted to binary format based on ASCII codes.

[pic]

Fig. 2 Speech steganography using DSSS with FFT domain

Then by using the pseudo noise and key with a gain factor, watermark message will be embedded into the FFT audio. Now, the audio will be integrated with both input audio and the binary message, which is known as stego audio. To reconstruct it and to extract the secret message applies the inverse process of embedding.

D. Wavelet Analysis

A wavelet technique i.e., variable windowing has been introduced to overcome the STFT drawbacks. Wavelet analysis allows the use of long time intervals where we want more precise low-frequency information, and shorter regions where we want high-frequency information. In fig5, it is shown that the comparison of FT, STFT and wavelet transform by considering an example input signal and how the analysis of transformation techniques will apply to get the frequency information of input signal. We can observe that in wavelet analysis the graphical representation shows that the wavelet has more number of features than the FT and STFT. Wavelet is also called as multi resolution analysis (MRA).

E. Discrete Wavelet Transform (DWT)

In spite of the fact that the discretized persistent wavelet transform empowers the calculation of the consistent wavelet transform by PCs, it is not a genuine discrete transform. In actuality, the wavelet arrangement is basically a tested adaptation of the CWT, and the information it gives is profoundly redundant to the extent the remaking of the sign is concerned. This redundancy, then again, requires a lot of calculation time and assets. The discrete wavelet transform (DWT), then again, gives adequate information both to analysis and blend of the first flag, with a critical diminishment in the calculation time. The DWT is impressively less demanding to execute when contrasted with the CWT. The essential concepts of the DWT will be presented in this area alongside its properties and the algorithms used to figure it.

III. PROPOSED METHODOLOGY

A. Translation Invariant Wavelet Transform (TIWT)

Given a signal s of length N, the first step of the TIWT produces, starting from s, two sets of coefficients: approximation coefficients cA1 and detail coefficients cD1. These vectors are obtained by convolving s with the low-pass filter Lo_D for approximation, and with the high-pass filter Hi_D for detail.

More precisely, the first step is

[pic]

Fig. 3 decomposition tree of translation invariant transform

NotecA1 and cD1 are of length N instead of N/2 as in the DWT case.The next step splits the approximation coefficients cA1 in two parts using the same scheme, but with modified filters obtained by up sampling the filters used for the previous step and replacing s by cA1.Then, the TIWT produces cA2 and cD2. More generally,

[pic]

Fig. 4 1-D translation invariant wavelet transform

B. Algorithm

In this section, the steganography scheme will be explained. Cover object used is the raw audio files like WAV files. Suppose we have byte-sequence information that will be inserted into the cover object, the byte-sequence will be converted into bit-sequence information. Then we represent these bits into such signal so that if the bit is 1 then the amplitude of the signal is 1, whereas if the bit is 0 then the amplitude of the signal is -1. As shown as follows:

[pic] (1)

Next, open the WAV files and obtain the signal’s amplitude data. The amplitude is represented as 16 bit signed integer value with a range of 215-1 to -215+1. So, for divide this amplitude with a value of 215-1 in order to obtain the range between 1 to -1. Then, by using FFT, the data will be transformed into frequency domain. Now, a random PN sequence will be generated with 1 or -1. If the chip rate of PN sequence is cr, and there are total of n signals for the information signal, then there must be [pic] sequences generated. We call the PN sequence P, then

[pic] (2)

Modulate each information signal with the PN sequence until cr times, by multiplying the value. It will produce a signal B which is the distributed signal of A and of course with length cr times its original length. Initially, spread the information in A to B as follows:

[pic] (3)

Next, modulate B and P and multiply it by a factor α. Then it will be injected into the cover-media. Suppose, the message that is injected is w, the cover is v and the stego object v’ i.e., in which we have both the message as well as cover-object. Therefore, it can be formulated as follows:

[pic] (4)

[pic] (5)

This scheme will generate noise. If the factor of the amplifier is too large, the noise is also large and may damage the cover-object. So, we should be careful in choose of the strength factor and chip-rate. The added signal will be a random signal due to the PN sequence effect which has generated previously. In order for the information to be retrieved, the receiver must generate the same PN sequence. Each cover object signal will be multiplied with the corresponding PN sequence, which can be shown as follows:

[pic]

If we look at the following terms:

[pic]

The value of these terms will be close to 0 for a large number of samples (large chip rate). This is because the random value of PN sequence causes the sum of the signal approaching 0 or a certain threshold value.

While the second term:

[pic]

The second term has interesting properties. Because the PN sequence has value 1 or -1, then the result of [pic] is 1.

Thus, the term can be simplified into:

[pic]

Because we have defined Bi has a value 1 or -1, then we simply conclude that if the term exceeds the value of zero, we assume that the information retrieved is 1 and if the value is less than zero, we assume that the information retrieved is 0. This is the reason we choose the domain of B and P. From the previous explanation, we can conclude that the value of [pic] must exceed a certain threshold value in order for a clear information retrieval.

IV. EXPERIMENTAL ANALYSIS

Simulation results have been done in MATLAB 2014a. We tested the proposed and existing methods for various audio samples with different cr values.

[pic]

(a)

[pic]

(b)

Fig. 5 Performance of FFT and proposed TIWT based Speech Steganography (a) Cover, Stego and Reconstructed speech signals obtained by FFT implementation (b) Cover, Stego and Reconstructed speech signals obtained by TIWT implementation

In fig5 (a) FFT based audio steganography has been shown, in which the stego audio is different from the original audio i.e., the unauthorized party can observe the difference in original and stego audio which in results insecure system. We have found that both original and stego audio looks like same in fig5 (b), which has achieved by our proposed TIWT approach, also compared the both existing and proposed schemes in fig5 (a) and (b).

[pic]

(a)

[pic]

(b)

[pic]

(c)

[pic]

(d)

[pic]

(e)

Fig. 6 user interface outputs of proposed TIWT based speech steganography (a) user interface (b) cover speech (c) secret message (d) embedding process/stego speech (e) extracted message and reconstructed speech

V. CONCLUSIONS

Implementation of novel user interface for spread spectrum representation-based speech steganography using translation invariant wavelet transform is done successfully. Obtained simulations proven that the proposed TIWT based speech steganography got superior performance over conventional steganography algorithms like FFT and DWT. This method proved that it is very robust against audio manipulation and very safe with the resulting noise is quite small. Also it reduces number of computations and does not use any complex equations. It is very simple and easy method to implement even in real time environment.

REFERENCES

1. Shouyuan Yang, Zanjie Song and Jong Hyuk Park “High capacity CDMA Watermarking Scheme based on orthogonal Pseudo random subspace projection”. International Conference on Multimedia and Ubiquitous Engineering, June 2011

2. Lionel Fillatre “Adaptive Steganalysis of Least Significant Bit Replacement in Grayscale Natural Images” IEEE Transactions on Signal Processing ,Vol. 60, No. 2, February 2012

3. R.R.Ahirwal, Deep chand Ahirwal and Jpgendar jain “A High Capacitive and Confidentiality based Image Steganography using Private Stego key” International coference on Information Science and applications, Feb 2010.

4. Rikzy M. Naguraha “Implementation of Direct sequence Spread Spectrum on Audio Data” International Conference on Informatics Engineering, June 2011.

5. Siwar Rekik, Driss Guerchi,Habib Hamam & Sid-Ahmed Selouani “Audio Steganography Coding Using the Discrete Wavelet Transforms”.International Journal of Computer Science and Security (IJCSS), Volume (6) : Issue (1) : 2012

6. Jie Chen, Jose Carlos, “A Spread Spectrum Representation Based FFT Domain Speech Steganography Method”, IEEE Transaction on Audio, Speech and Language letters, Vol. 23, No. 1, 2015.

-----------------------

ISSN NO: 0745X6999

JOURNAL OF RESOURCE MANAGEMENT AND TECHNOLOGY

Vol , Issue , month / year



Page No:

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download