An Overview on Audio Steganography Techniques Vol6 No13 Binder1... · 2012-07-31 · lower payload of the watermarking methods compared to steganographic algorithms [35, 42, 47, 63].

An Overview on Audio Steganography Techniques

1Mazdak Zamani, 2Azizah Bt Abdul Manaf, 3Shahidan M. Abdullah 1, 2,3Advanced Informatics School, Universiti Teknologi Malaysia,

[email protected], [email protected], 3 [email protected]

Abstract Steganography is a form of security technique through obscurity; the science and art of hiding the

existence of a message between sender and intended recipient. Steganography has been used to hide secret messages in various types of files, including digital images, audio and video. The three most important parameters for audio steganography are imperceptibility, payload (bit rate or capacity), and robustness. Any technique which tries to improve the payload or robustness should preserve imperceptibility. The noise which is introduced due to bit modification would limit payload. This paper presents a categorization of information hiding techniques and overviews those techniques that intend to improve payload and imperceptibility.

Keywords: Artificial Intelligence, Multimedia Security, Steganography, Watermarking

1. Introduction

Steganography and watermarking techniques embed information in a media in a transparent manner.

Steganography is a technique for covert information, but watermarking may not hide the existence of the message from third persons. Watermarking usually requires robustness to withstand against attacks intended to remove or destroy the hidden message from the watermarked media as well as preserving the carrier signal quality. This makes watermarking appropriate for those applications where the knowledge of a hidden message leads to a potential danger of manipulation [4, 27, 33, 58,59].

The most well-known examples of steganography go back to ancient times when Histiaus shaved his slave’s head, and then he tattooed a message on his scalp. After that his hair had re-grown the tattooed message was disappeared. He was going to call his men to attack to the Persians [28, 39, 62].

Steganography is the study of methods for hiding the existence of secondary information in the presence of primary information in a way which neither affects on the size nor results in perceptual distortion. The secondary information is referred to as hidden message, hidden file or hidden information while primary information is referred to as carrier, host or original signal, before embedding and stego signal, file, bit stream or sequence, after embedding [2, 6, 20, 60, 61, 64].

Watermarking techniques are principally context-specific, that means, the algorithms must be designed regarding the media type of the data to be watermarked. Therefore, watermarking indicates a specific application of steganographic techniques. Specifically, the additional requirement for robustness of digital watermarks against attacks or manipulations during the data processing entails a lower payload of the watermarking methods compared to steganographic algorithms [35, 42, 47, 63].

1.1. Overview of the human auditory system

Since dynamic range of the HAS is wider in comparison to human visual system (HVS), hiding data

in audio is more difficult in comparison with the hiding data in images or video. Digital audio technologies like the development of MPEG rely on the detailed knowledge of the human auditory system.

The relevant information is not only limited by the ability of the ear to hear frequencies in a band between 20 Hz and 20 kHz and the dynamic range of over 96 dB, but the physical phenomenon of different frequencies and their corresponding process in HAS is also important to consider in depth the correlation between hearing sensations and acoustical stimuli. A phenomenon is that the sensitivity of human ear to changes in louder audios is more compared to in quieter audios [26, 31, 34].

1.1.1. Listening test

An Overview on Audio Steganography Techniques Mazdak Zamani, Azizah Bt Abdul Manaf, Shahidan M. Abdullah

International Journal of Digital Content Technology and its Applications(JDCTA) Volume6,Number13,July 2012 doi:10.4156/jdcta.vol6.issue13.13

107

Listening test is an important evaluation mechanism in audio steganography. Listening test is also important to find out how that is being done in the science world. An overview of the properties of the Human Auditory System (HAS) is given.

1.1.1.1. ABX listening test

A listening test is usually performed to evaluate the quality of audio signals. An ABX listening test

is a common way to evaluate the quality of data hiding method in audio signals. Three different audio clips are presented to listeners in ABX test: selection A which consists of

original audio, selection B which consists of the watermarked audio and X which consists of either the watermarked or original audio, drawn at random. Then listeners are asked to make decision whether selection X equals to A or B. The number of correct answers verifies whether the stego or watermarked audio and the original audio are perceptually different and, thus, the data hiding algorithm will be declared as “perceptible.” Otherwise, that will be confirmed as imperceptible or transparent [36-38].

1.1.1.2. Distinguishing listening test

Tiberiu et al. [54] used a modulation with an adaptive carrier frequency for a procedure based on

frequently substitution, depending on the original signal features. Several audio files with 44.1 KHz sampling frequency were used for the test. Several listeners were presented to distinguish between stego signals and original signal by listening [54].

2. Information Hiding Techniques

In this section all techniques of data hiding are classified. Referring to the block diagram in Figure 1,

the first class of information hiding methods is watermarking. The aim of watermarking is to embed copyright information within a host data [45, 48, 49].

Figure 1. Classification of Information Hiding

2.1. Watermarking methods


108

One type of watermarking is fragile watermarking which aims to hide information in a fragile way such that in case of any attack, it is possible confidently, to prove that the data has been altered. Thus, the objective is to propose a hiding system that is able to detect any subsequent alteration. Another type of watermarking is robust watermarking. This is unlike fragile watermarking, as long as the quality of the attacked object is “acceptable” for all practical purposes. Attackers should not be able to degrade the quality of the watermark detector significantly [15, 21, 41].

In blind watermarking, the assumption is that the original host is not present at the receiver side due to the practical limitations of the scenarios where watermarking could be useful. If the existence of the host data at the receiver end is assumed, that would be called non-blind watermarking or private watermarking.

In fingerprinting, the copyrighted data are distributed to different users who possibly have a different identification code [17, 46, 56].

2.2. Steganography methods

Steganography establishes a covert communication channel between two parties. The existence of

this channel should be unknown to possible attackers [7, 13, 57]. A small category of digital steganography is embedding in text. In this method a message is hidden

in the blank spaces between words. As a result sender has to make sure the text is sending is considerably longer than the message to be hidden [19, 53].

Since most techniques in the field of digital steganography are categorized in the media steganography, media steganography is usually referred as digital steganography. Every technique of digital steganography can be categorized into two general categorizes: Spatial techniques and temporal techniques.

2.2.1. Spatial domain methods

Spatial techniques category contains those techniques which deal with bits stream to hide message.

Mostly, these techniques change values of bits in audio file in order to embed the message and do not deal with signal processing concepts [8, 12].

Data injection embeds the secret message directly in the host medium. The message is hid in those sections of a file where will be ignored by the processing application. For example, written data after End of File (EOF) marker is nonexistent as far as meaningful content is concerned. So that part of file could be used to embed message data for those file types and programs which an EOF marker signifies the end of a file [1, 24, 30].

Substitution-based algorithm replaces the least significant bits of the original file content with message data in a manner that the least amount of distortion is caused. During execution of the algorithm, file size is kept the same; however the payload is limited to the amount of insignificant bits in the file [29, 32].

Audio Least Significant Bit (LSB) steganography takes advantage of the quantization error that usually derives from the task of digitizing the audio signal. As the name states, the information is encoded into the right most bits per samples or least significant bits from audio data. Ideally, the embedding capacity of an audio file with this method is 1 kbps per 1 kHz of sampled data. That is, if a file is sampled at 44 kHz then it is possible to embed 44 kilobits on each second of audio. In exchange for that high channel payload, audible noise will be introduced [3, 25].

Image Least Significant Bit steganography alters insignificant bits of host image. With a 24-bit image, 3 bits per pixel could be one can stored by altering one bit from each blue, green and color which are defined by 3 bytes. In average, to hide a message just fifty percent of image bits will be changed. Regarding the 256 level of intensity for a primary color, with a bit altering in a pixel, only a small difference in the amount of color intensity will be resulted. The human eye is not able to recognize the difference – and as a result the message can be successfully hidden [29, 33]. 2.2.2. Temporal domain methods

Temporal domain techniques usually use psychoacoustics model to perceptually weight the

introduced noise. Many techniques that are mostly robust have been categorized in this category.


109

Transformation domain techniques are a large category of temporal techniques. Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) hide the messages in the significant areas of the cover file. Discrete Cosine Transform (DCT) and Discrete Fourier Transform (DFT) are so similar [3, 28, 55].

Short Time Fourier Transform (STFT) is created by modifying the Fourier transform to take advantage of both time and frequency information. Continuous Wavelet Transform (CWT) and Discrete Wavelet Transform (DWT) are different types of wavelet transforms [27, 37].

In Frequency masking, by a simultaneously appearing stronger signal, a low level signal is possibly made inaudible.

Dither embedding is another category of audio based techniques. Dither is a noise signal that is added to the input audio signal to provide better sampling of that input when digitizing the signal. As a result, distortion is practically eliminated, at the cost of an increased noise floor. To implement dithering, a noise signal is added to the input audio signal with a known probability distribution, such as Gaussian or triangular. In the particular case of dithering for embedding, the message or watermark is used to modulate the dither signal. The host signal (or original audio file) is quantized using an associated dither quantizer [31, 38].

Echo hiding embeds data into a host file by additional echoes to create stego file. Naturally an echo adds a resonance to a host audio. After adding the echo, stego audio keeps the same perceptual characteristics and statistical. Because the delay between host and stego audio is small enough, HAS perceives the echo as an added resonance [29].

Phase coding (or phase distortion) substitutes the phase of the original audio signal with one of two reference phases, each one encoding a bit of information. Phase coding algorithms use the fact that HAS is not sensitive to a constant relative phase shift in a stationary audio signal, rather than masking properties of the human auditory system.

There are two basic approaches to spread spectrum techniques: direct sequence and frequency hopping. In both of these approaches the idea is to spread the embedding data across a large frequency band, namely the entire audible spectrum. In the case of direct sequence, the cover signal is modulated by the watermark message and a pseudorandom noise sequence, which has a wide frequency spectrum. In the case of frequency hopping, the cover frequency is altered using a random process, thus describing a wide range of frequency values [28, 30].

2.2.3. Linguistic steganography methods

Second category of steganography is linguistic steganography and attracts the attention of

researchers during last decades. Our language is in itself a code and to anyone whom has not learned it appears incomprehensible. Linguistic steganography has two categories, open ciphers and semagrams. Semagrams hides information by using symbols and signs [25-27].

A visual semagram is agreed codes that are going to be transmitted by placing an item in a specific location on your desk or waving your hand or so on. These signs take advantage of the normality of nowadays world and as a result are difficult to be detected. Textual semagrams hides message into texts by some arranged works like capitalized letters or blank spaces in between words [32-36].

Open ciphers codes the message in a legitimate piece of text in a manner that is not immediately observable. The most obvious use of open ciphers occurs in the use of code words where individual characters, words are mapped onto entities of the carrier signal. Jargon is first category of open ciphers that is done by prearranged meanings or underground terminology to hide the message [39].

Misspellings are another category of open ciphers and take advantage of the number of variations of the spelling of a word. The meaning of a word with some incredibly misspelling can be retained to bypass some electronic filters which are programmed to react to certain words. Phonetics is another category of open ciphers and is used when a local language or the keywords of a popular second language is applied to confuse a filtering systems. Covered ciphers hide a message openly so that only one can recover the message who knows the message exists or how that can be decrypted [3, 34, 49].

Grille ciphers, is based on the imposition of a grid known only to the communicating parties onto a message consisting of characters or words commonly attributed to Girolamo Cardano and reading the elements left uncovered by the grille in a predefined order. In Null cipher, the payload message is embedded as plain text within the carrier message. The communicating parties prearrange a set of rules that specify the extraction of the payload message [48].


110

Last category of steganography is physical steganography that involves the use of technical means to conceal the existence of a message using physical or chemical means [35, 36].

3. Literature Review

In this section, similar research related to our work is presented. These literatures are categorized

regarding the aspects which are to be compared with our work.

3.1. Improving imperceptibility in PSNR

Significant improvement in PSNR in some related works are presented in this section. Following are some related works with their results.

Ji et al. [18] enhanced PSNR value for LSB steganography with a secluded statistic peculiarity. They found the best mapping function between host and secret image blocks at global scope and as a result could minimize the degradation of the stego. Table 1 shows the comparison of their proposed method and some other methods by embedding 256×256 pixel size images as messages into 512×512 pixel size images as hosts [18].

As Table 2 shows, their work improved the PSNR by 1.63 dB compared to the simple LSB. Regarding the payload of their proposed method is quite high; the amount of improvement is acceptable. The problem of their proposed method is its imperceptibility, because that is about 45 dB, which is not high enough for some application.

Wang et al. [54] developed a genetic algorithm to hide message data in the K-rightmost LSBs of the host image. However the computation time to find the optimal result might be huge when K is large. Also an improved hiding technique to obtain a high-quality embedding is developed based on the concept of perceptual modeling. They tested their experiment on the 8-bit images with 256 gray levels. Figure 2 shows the result of embedding image. As can be seen three different operations, including the simple substitution, the optimal substitution, and the proposed GA approach are conducted. Also the PSNR obtained by the GA approach is very high, which makes the quality of the embedding result acceptable [54].

As Table 3 shows, their work improved the PSNR by 0.55 dB in average compared to the simple LSB. Regarding their proposed method does not provide high capacity; the amount of improvement is relatively low.

Ming et al. [40] applied the LSB substitution and genetic algorithm (GA) to improve the image quality of the stego-image, a developed two different optimal substitution strategies: one is the global optimal substitution strategy and the other is the local optimal substitution strategy [40]. The PSNR values of the four methods, including Wang et al.’s optimal substitution method, the LSB substitution method, their global method, and their local method, are listed in Table 4.

Table 1. PSNR Comparison [18]

BPCS MODULE OPAP LSB OLSB ProposedMethod43.19 32.32 46.36 44.15 44.16 45.78

Table 2. Ji's method performance [18]

LSB Ji'smethodPSNR 44.15 45.78Amountofimprovement(indB) 1.63


111

Figure 2. The result of embedding [54]

Table 3. Wang's method performance [54]

SimpleLSBWang'smethod

Message1 43.96 44.53Message2 44.14 44.28Message3 43.96 44.9Average(inPSNR) 44.02 44.57Amountofimprovement(indB) 0.55

Table 4. The test results [40]

Table 5. Ming's method performance [40]

SimpleLSB Wu'smethod

Message1 32.43 33.16Message2 32.28 32.67Message3 32.44 32.98Average(inPSNR) 32.38 32.94

Amountofimprovement(indB) 0.56

As Table 5 shows, their work improved the PSNR by 0.56 dB in average compared to the simple

LSB. However the payload of their proposed method is quite high; but the amount of improvement is low. Also, another problem of their proposed method is its imperceptibility, because that is about 33 dB, which is not high enough for some application.

Shih et al. [50] applied genetic algorithm to come up with a steganography system that breaks the steganalytic systems. They used an approach, differential evolution (DE), to enhance the performance of the steganography system. Experimental results show that the PSNR of the stego-image is improved by DE application. Many cover images with a 256 × 256 size and 10 secret messages with a 64 × 64 size were tested. Table 6 shows the result of embedding for some different iteration [50].

However their proposed technique can provide an acceptable capacity with an imperceptible level of PSNR, but that needs 100 iterations which is time-consuming and if the size of host file is large, this method may not be applicable.

Liu et al. [22] proposed a variable depth substitution technique for data hiding. Table 7 shows the result of embedding in which the image quality of the stego-image hidden by this method improves from 0db to 4db against simple LSB (Less significant Bit) substitution method. Chosen cover image size is 128 by 128 [22].

As Table 8 shows, their work improves the PSNR by 2.33 dB in average compared to the simple LSB. However amount of improvement is acceptable, but the payload is low.


112

Table 6. Comparison of GA and DE algorithms [50]

Table 7. The average PSNR with 100 random message embedding [22] Method SimpleLSB VariableLSB3k 44.16 46.355k 37.93 40.317k 31.87 34.29

Harsh et al. [16] proposed a novel audio watermarking method in which audio file is partitioned into

frames of 90 ms duration. Then a random sample is selected using the key and watermark is added into the lowest bit of the selected sample by the following procedure: Reading one byte of the message stream, for each bit in (message): reading a byte from the key, skip a couple of samples, read one sample from the wave Stream, getting the next bit from the current message byte, and placing it in the last bit of the sample. Five different types of 16-bit mono audio signals sampled at 44.1 kHz were selected to evaluate the performance of the proposed algorithm [16].

The experimental results for SNR (Signal to Noise Ratio), PSNR (Peak Signal to Noise Ratio) and BER (Bit Error Rate) are given in Figure 3. Regarding the level of obtained PSNR, their proposed technique is imperceptible, but since the algorithm skips a couple of samples to embed next bit, the payload is very low.

Andres et al. [5] tabulated some different metrics like PSNR found on the literature on Table 9. Even though the acceptable amount of noise depends on both the data hiding application and the properties of the original audio host, it could be expected to get perceptible noise distortion for SNR values of 35 dB [5].

Table 8. Liu's method performance [22]

SimpleLSB Liu'smethod

Message1 44.16 46.35Message2 37.93 40.31Message3 31.87 34.29Average(inPSNR) 37.99 40.32Amountofimprovement(indB) 2.33


113

Figure 3. Watermarking algorithm’s performance [16]

Table 9. Common difference distortion metrics [5]

3.2. Improving payload in bps

This is also important to find out how many bits are going to be embedded in each sample or find

out how much payload can be achieved by improving the level of PSNR in other works. Following are some related works with their results.

Liu et al. [23] transformed data hiding problem into finding the representative element in specific equivalence class to propose a data hiding strategy based on equivalence class. Then for a LSB hiding scheme they used minimizing the distortion in the equivalence class (MDEC). In fact, there is a tradeoff between length of information and distortion. The performance was measured by worst mean-square-error (WMSE) and worst PSNR (WPSNR) [23].

In Figure 4, the WPSNR is indicated by y-axis. The problem of their proposed technique is when the payload is high, where the PSNR comes below 40 dB.


114

Figure 4. WPSNR under different hiding schemes [23]

Figure 5. Relation between SNR and hiding capacity [51]

Shirali et al. [51] presented a high capacity and low error rate steganography with adaptive wavelet

domain. They tested 4 sample wav files with 16 bit per sample, 44100 Hz sample rate and 10 second duration from four different music genres. In Figure 5 the relation between SNR of the stego audio and the payload is shown. Also it is shown that even with a payload up to 200 kbps the stego audio will be imperceptible from original audio. For different types of music, the SNR is almost the same. To do listening test, both original and stego audio were played for testers involving three men and a woman [51]. They were asked to select the stego audio. When the bit per sample rate is not high, the PSNR is acceptable, but the problem of their proposed technique is when the payload is high, where the PSNR comes below 40 dB.

Sos et al. [52] used an algorithm based on classical unitary transforms with quantization in the transform domain to embed the secure data. Table 10 provides the analysis of sample audio signals. As can be seen, the Maudio measure can be a method to select an audio clip in order to have the least distortion [52].

Regarding the second column of the table, the problem of their proposed technique is its payload which is 1 bit per 63 bits.

Chang et al. [9] proposed Vector Quantization method for steganography to embed into binary images and improve the quality. They used a new S-tree and applied the genetic k-means for reducing the replacement distortion. That is shown in Table 11 with the PSNR value between their methods in comparison with other schemes after the embedding of the fixed size random data (Kbits) [9].


115

Table 10. Quantization in frequency domain [52]

Table 11. Performance comparisons in PSNR among various methods [9]

The results shows when the same random data size is embedded, the proposed method has

performed constantly better than Du and Hsu’s method. Regarding the last column of the table, the problem of their proposed technique is its imperceptibility which is below 30 dB.

Cvejic et al. [10] proposed embedding process in the wavelet domain to increase the payload of the classic LSB insertion. The algorithm embeds message bits into wavelet domain of audio file by modifying LSB values of wavelet coefficients. Proposed algorithm improves the performance of classical LSB algorithm by 150-200 kbps of hidden data for the same SNR values. Modifying one LSB gives a payload of 44.1 kbps, if 44100 Hz sampled 16 bits per sample audio clips are used. The SNR value and capacity of hidden data for a tested audio file is shown in Figure 6 [10].

Obviously has higher SNR value (around 20 dB) for the same capacity of hidden data for wavelet LSB insertion (upper curve) is higher than the classic time domain insertion. When the bit per sample rate is not high, the PSNR is acceptable, but the problem of their proposed technique is when the payload is high, where the PSNR comes below 40 dB.

Nedeljko et al. [44] proposed a modification to standard LSB algorithm to make it capable to embed four bits per sample. The algorithm uses minimum error replacement method for LSB adjustment to decrease the SNR value. SNR value of audio with hidden data and the original audio piece against capacity of channel for additional data is shown Figure 7. Upper curve depicts SNR values for the modified LSB algorithm, for data hiding in two, three and four LSBs used. Curve below depicts SNR values for standard LSB algorithm for different hidden data channel capacity values. It can be seen that depending on the number of LSBs used for data hiding, the algorithm can outperform standard LSB insertion algorithm for 2.7 to 4.0 decibels. The performance of the algorithm in the case of four LSBs (176.4 kbps) is close to the standard method in the case of three LSBs (132.3kbps) [44].

Actually this technique has improved the imperceptibility in an acceptable level, but there still a problem when the payload needs to be high.


116

Figure 6. Capacity of hidden data channel versus SNR [10]

Figure 7. Capacity of hidden data versus SNR value [44]

Table 12. Performance of the proposed scheme [11]

Delforouzi et al. [11] proposed a method for audio steganography. In the proposed method, in the

integer wavelet domain, encrypted covert data is embedded into the coefficients of the host audio. Eight music clips and human voices that are sampled at 44.1 KHz with a length of about 5–100 seconds and quantized by 16 bits were tested with the proposed algorithm. Table 12 shows analyzed results of the performance of the algorithm for SNR and BER [11].

However the imperceptibility of their proposed technique is not high, but regarding the payload is high, that is acceptable. The problem is their proposed technique is not capable or not tested for less payload and higher quality in imperceptibility.


117

3.3. Improving efficiency of algorithm

This is also important to compare the efficiency of those algorithms which are used in related works and their results.

Foo et al. [14] reported the robustness and performance of One Bit per Sample (OBS) approach one method of audio watermarking is to embed one bit of the watermark in each sample of host signal. The watermark bit can be embedded by replacing the least significant bit (LSB) or higher bit of host signals. To minimize the change in sampling value that results from a change in the bit at a higher bit, other bits have to be changed [14].

For example, if one original sample value is 00000000000110002 =2410, and the watermark bit is ‘0’ which is to be embedded at the 4th bit. Without adjusting the other bits, the binary representation of the watermarked sample value is 00000000000100002=1610. This value is far from the original value of 24; hence some other bits may have to be changed. One change is as follows: 0000000000101112=2310. The watermark bit remains at the 4th bit, while the value of the sample is now closer to 24. Let a 16-bit binary sample of the signal be represented by a16, a15, …,a1, where a1 is the bit at the 1st bit (LSB); a2 represents the bit at the 2nd bit and so on. If the ith bit is chosen as the watermark embedded bit and if ai of the sample is already equal to the watermark bit, no action is taken. If ai of the sample is not equal to the watermark bit to be embedded, other bits of the sample are to be changed to minimize the change in the value of the sample [14].

In order to investigate the effects of addition of white noise, additional different amount of white noise are added to those 5 original watermarked audio signals and the resulting SNRs are computed. All resulting SNRs are given in Table 13. Results of experiments show that when no additional noise is added, the SNR of whole watermarked signal is only affected by the watermark bits that change the values of the samples. There is a drop of 25.8 dB (67.94dB-42.14dB) in the SNR obtained when watermark bits are embedded at 5th bit as compared with the SNR for embedding at 1st bit. The SNR of the watermarked signal using 5th bit embedding is only 42.14 dB. Informal listening test reveals that the quality of the watermarked signal is significantly affected. Hence they limit their experiments to embedding at 5th bit [14].

Nedeljko et al. [43] presented a high payload LSB audio watermarking method to reduce the distortion of embedding in audio. They developed a method that shifts the limit for transparent data hiding in audio from the fourth LSB bit to the sixth LSB bit. As the algorithm given in Appendix B shows, in the embedding algorithm, the (i+1)th LSB bit (bit ai) is first modified by insertion of the present message bit. Bits of watermarked audio are represented by Underlined bits (ai). In case that the bit ai are already at a correct value there is no need for modifying at all and no action is taken with that signal sample [43].

They tested proposed algorithm on 10 audio sequences from different music styles including pop, rock, techno, and jazz and were 44.1 kHz sampled mono audio files, represented by 16 bits per sample, with duration from 10 to 15 seconds. Figure 8 shows the SNR values for the standard method that performs embedding in the 4th LSB bit and the proposed method that performs embedding in the 4th, 5th and 6th LSB bit. Also the algorithm outperforms standard LSB insertion algorithm. Both 6th LSB bit using the proposed method and 4th LSB bit using the standard method obtain similar SNR values [43].

Table 13. Effect of adding white noise [14]


118

Figure 8. SNR values comparison [43]

4. Acknowledgment This work is part of research done with support from Advanced Informatics School (AIS), Universiti Teknologi Malaysia (UTM). 5. References [1] Abduljabbar Shaamala, Shahidan M. Abdullah and Azizah A. Manaf, Study of the effect DCT and

DWT domains on the imperceptibility and robustness of Genetic watermarking, IJCSI International Journal of Computer Science Issues, Vol. 8, Issue 5, No 2, September 2011

[2] Abduljabbar Shaamala, Shahidan M. Abdullah and Azizah A. Manaf, The Effect of DCT and DWT Domains on the Robustness of Genetic Watermarking ICIEIS 2011, Part I, CCIS 251, pp. 310–318, 2011, Springer.

[3] Akram M. Zeki, Azizah A. Manaf, Adamu A. Ibrahim and Mazdak Zamani. A Robust Watermark Embedding in Smooth Areas. Research Journal of Information Technology. Year: 2011. Volume: 3. Issue: 2. Page No.: 123-131. DOI: 10.3923/rjit.2011.123.131. ISSN: 2151-7959.

[4] Akram M. Zeki, Azizah A. Manaf, and Mazdak Zamani. Bit-Plane Model: Theory and Implementation. Engineering Conference 2010 (EnCon2010). 14-16 April 2010. Kuching, Sarawak, Malaysia.

[5] Andres G. (2002). Measuring and Evaluating Digital Watermarks in Audio Files. Washington DC. 2002.

[6] Anguraj S., Dr.Shantharajah S.P., Arul Murugan R., Balaji E., Maneesh R., Prasath S. A Fusion of A-B MAP Cipher and ASET Algorithms for the Enhanced Security and Robustness in Audio Steganography. IEEE-International Conference on Recent Trends in Information Technology, ICRTIT 2011. MIT, Anna University, Chennai. June 3-5, 2011.

[7] Asha Vijayan, Amrita Vishwa Vidyapeetham. A Genetic Algorithm Approach for Audio Steganography. Computer Science Journals.

[8] Awodele Oludele. Artificial Intelligence and Security. International Journal of Computational Intelligence and Information Security Vo. 1 No. 3, May 2010.

[9] Chang, C., Shen, W., Wang, H. (2007). Using Counter-Propagation Neural Network For Robust Digital Audio Watermarking In DWT Domain. IEEE International Conference on Systems, Man and Cybernetics, Volume (2), 1214-1219.

[10] Cvejic, N., Seppanen, T. (2002). A Wavelet Domain LSB Insertion Algorithm for High Payload Audio Steganography. Digital Signal Processing Workshop. 13-16 October 2002. 53-55.

[11] Delforouzi, A., Pooyan, M. (2008). Adaptive Digital Audio Steganography Based On Integer Wavelet Transform. Circuits, Systems, and Signal Processing. Volume (27). 247-259.

[12] Ei Thin Su. Robust Data Embedding Based Probabilistic Global Search in MDCT Domain. Informatics Engineering and Information Science. Communications in Computer and Information Science, 2011, Volume 251, Part 2, 290-300, DOI: 10.1007/978-3-642-25327-0_25.

[13] Fadhel, N.F., Wills, G.B., Argles, D. Transparent authentication in e-leaming. International Conference on Information Society, i-Society. 2011, art. No. 5978465, pp. 336-342.


119

[14] Foo, S., W., Dong, Q. (2009). Robustness of One Bit Per Sample Audio Watermarking. IEEE International Symposium on Circuits and Systems, 932-935.

[15] H.B.Kekre, Archana Athawale, Swarnalata Rao, Uttara Athawale. Information Hiding in Audio Signals. International Journal of Computer Applications. Number 9 - Article 4. Year of Publication: 2010. DOI: 10.5120/1278-1623.

[16] Harsh V., Ramanpreet K., Raman K. (2009). Random Sample Audio Watermarking Algorithm for Compressed Wave Files. International Journal of Computer Science and Network Security, Volume (9), November 2009.

[17] Hui Tian, Ke Zhou, Jing Lu. A VoIP-based Covert Communication Scheme Using Compounded Pseudorandom Sequence. International Journal of Advancements in Computing Technology (IJACT). Volume4, Number1, January 2012.

[18] Ji, R., Yao, H., Liu, S., Wang, L. (2006). Genetic Algorithm Based Optimal Block Mapping Method for LSB Substitution. International Conference on Intelligent Information Hiding and Multimedia Signal Processing. 215-218.

[19] Kekre, H.B., Athawale, A., Rao, B.S. & Athawale, U. 2010, "Increasing the capacity of the cover audio signal by using multiple LSBs for information hiding". 3rd International Conference on Emerging Trends in Engineering and Technology, ICETET 2010. 2010, Article number 5698319, Pages 196-201.

[20] Krishna Bhowal, Anindya Jyoti Pal, Geetam S. Tomar, P.P. Sarkar. Audio Steganography Using GA. 2010 International Conference on Computational Intelligence and Communication Networks. Bhopal, India. November 26-November 28. ISBN: 978-0-7695-4254-6.

[21] Krishna Bhowal, Debnath Bhattacharyya, Anindya Jyoti Pal and Tai-Hoon Kim. A GA based audio steganography with enhanced security. Journal Name: Telecommunication Systems. DOI: 10.1007/s11235-011-9542-0.

[22] Liu, S., Chen, T., Yao, H., Gao, W. (2004). A Variable Depth LSB Data Hiding Technique in Images. International Conference on Machine Learning and Cybernetics. Volume (7), 3990-3994.

[23] Liu, S., Yao, H., Gao, W., Yang, D. (2007). Minimizing The Distortion Spatial Data Hiding Based on Equivalence Class. Lecture Notes in Computer Science. Volume (4681), 667-678.

[24] Mayank Srivastava, Mohd Qasim Rafiq. A Novel Approach to Secure Communication Using Audio Steganography. Advanced Materials Research Vols. 403-408 (2012) pp 963-969. November, 2011.

[25] Mazdak Zamani, and Azizah Abdul Manaf. “Azizah’s Formula to Measure the Efficiency of Steganography Techniques”. 2nd International Conference on Information and Multimedia Technology (ICIMT 2010). December 28-30, 2010. Hong Kong, China.

[26] Mazdak Zamani, and Azizah Abdul Manaf. “Mazdak’s Method to Estimate the PSNR of Audio Steganography Techniques”. International Conference on Computer and Computational Intelligence (ICCCI 2010). December 25-26, 2010. Nanning, China.

[27] Mazdak Zamani, Azizah Abdul Manaf, and Rabiah Ahmad. “Current Problems of Substitution Techniques of Audio Steganography”. The 2009 International Conference on Artificial Intelligence and Pattern Recognition. ISBN: 978-1-60651-007-0. Pages 154-160. 13-16 July 2009. Orlando, Florida, USA.

[28] Mazdak Zamani, Azizah Abdul Manaf, and Rabiah Ahmad. “Knots of Substitution Techniques of Audio Steganography”. The 2009 International Conference on Telecom Technology and Applications. ISBN: 978-1-84626-017-9. Pages 415-419. June 6-8, 2009. Manila, Philippines.

[29] Mazdak Zamani, Azizah Abdul Manaf, Hossein Rouhani Zeidanloo, and Saman Shojae Chaeikar. “Genetic Subtitution-Based Audio Steganography for High-Capacity Applications”. International Journal for Internet Technology and Secured Transactions (IJITST). Pages 97-110. ISSN 1748-569X (Print), ISSN: 1748 - 5703 (Online). Volume 3 Issue 1, April 2011. Inderscience Publishers, Geneva, Switzerland. Doi:10.1504/IJITST.2011.039681.

[30] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, Akram Zeki, and Shahidan Abdullah. “Genetic Algorithm as an Approach to Resolve the Problems of Substitution Techniques of Audio Steganography”. The 2009 International Conference on Genetic and Evolutionary Methods. 13-16 July 2009. ISBN: 1-60132-106-6 and 1-60132-092-2. Pages 170-175. Las Vegas, Nevada, USA.

[31] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, Akram Zeki, and Shahidan Abdullah. “A Genetic-Algorithm-Based Approach for Audio Steganography”. International Conference on


120

Communities and Communications. World Academy of Science, Engineering and Technology 54 2009. ISSN: 2070-3740. Pages: 359-363. 24-26 June 2009. Paris, France.

[32] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, Akram Zeki, and Pritheega Magalingam. “A Novel Approach for Audio Watermarking”. Fifth International Conference on Information Assurance and Security. ISBN: 978-0-7695-3744-3. Pages 83-86. 18-20, August 2009. Xi'an, China.

[33] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, and Akram Zeki. “An Approach to Improve the Robustness of Substitution Techniques of Audio Steganography”. 2nd IEEE International Conference on Computer Science and Information Technology 2009. ISBN: 978-1-4244-4520-2 and 978-1-4244-4519-6. IEEE Catalog Number: CFP0957E-CDR. Library of Congress: 2009903791. Volume 2. Pages 5-9. 8 - 11 August 2009. Beijing, China.

[34] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, Farhang Jaryani, Hamed Taherdoost, Saman Shojae Chaeikar, and Hossein Rouhani Zeidanloo. “A Novel Approach for Genetic Audio Watermarking”. Journal of Information Assurance and Security 5 (2010). ISSN: 1554-1010. Pages 102-111.

[35] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, Farhang Jaryani, Hamed Taherdoost, Saman Shojae Chaeikar, and Hossein Rouhani Zeidanloo. “Genetic Audio Steganography”. International Journal on Recent Trends in Engineering & Technology [IJRTET], Volume 3, Issue 2, 2010. ISSN: 2158-5563. Pages 89-91.

[36] Mazdak Zamani, Azizah Abdul Manaf, Rabiah Ahmad, Farhang Jaryani, Hamed Taherdoost, and Akram Zeki. “A Secure Audio Steganography Approach”. The 4th International Conference for Internet Technology and Secured Transactions. ISBN: 978-0-9564263-1-4. Pages 501-506. 9-12 November 2009. London, UK.

[37] Mazdak Zamani, Hamed Taherdoost, Azizah A. Manaf, Rabiah B. Ahmad, and Akram M. Zeki. “An Artificial-Intelligence-Based Approach for Audio Steganography”. MASAUM Journal of Open Probelms in Science and Engineering (MJOPSE). Volume: 1 Issue: 1 Month: October 2009. ISSN: 2077-1118. Pages 64-68.

[38] Mazdak Zamani, Hamed Taherdoost, Azizah Abdul Manaf, Rabiah Ahmad, and Akram Zeki. “Robust Audio Steganography via Genetic Algorithm”. Third International Conference on Information & Communication Technologies ICICT2009. ISBN: 9781424446087. Pages 149 - 153. 15-16 August 2009. Karachi, Pakistan.

[39] Mazdak Zamani, Azizah Bt Abdul Manaf, Rabiah Bt Ahmad, Farhang Jaryani, Saman Shojae Chaeikar and Hossein Rouhani Zeidanloo. “Genetic Audio Watermarking”. International Conference on Recent Trends in Business Administration and Information Processing. March 26-27, 2010. Trivandrum, Kerala, India.

[40] Ming W., Min-Hui L. and Chin-Chen C. (2004). A LSB Substitution Oriented Image Hiding Strategy Using Genetic Algorithms. Lecture Notes in Computer Science. Volume (3309), 219-229.

[41] Muhammad Asad, Junaid Gilani, Adnan Khalid. An Enhanced Least Significant Bit Modification Technique for Audio Steganography. 2011 International Conference on Computer Networks and Information Technology (ICCNIT). Abbottabad. 11-13 July 2011. 143 – 147. ISSN: 2223-6317. INSPEC Accession Number: 12243911.

[42] Najaf Torkaman, M.R., Nikfard, P., Sadat Kazazi, N., Abbasy, M.R. & Tabatabaiee, S.F. 2011, Improving hybrid cryptosystems with DNA steganography. Communications in Computer and Information Science. Volume 194 CCIS, 2011, Pages 42-52.

[43] Nedeljko C., Seppäne T. (2005). Increasing Robustness of LSB Audio Steganography by Reduced Distortion LSB Coding. Journal of Universal Computer Science. Volume (11), 56-65.

[44] Nedeljko C., Seppanen, T. (2002). Increasing the Capacity of LSB-Based Audio Steganography. IEEE Workshop on Multimedia. Signal Processing. 336-338, 2002.

[45] Nuha Omran Abokhdair, Azizah Bt Abdul Manaf, Mazdak Zamani. Integration of Chaotic Map and Confusion Technique for Color Medical Image Encryption. 6th International Conference on Digital Content, Multimedia Technology and its Applications (IDC2010). 16-18 August 2010. Seoul, Korea.

[46] R. Darsana and Asha Vijayan. Audio Steganography Using Modified LSB and PVD. Trends in Network And Communications. Communications in Computer and Information Science, 2011, Volume 197, Part 1, 11-20, DOI: 10.1007/978-3-642-22543-7_2.


121

[47] R. Darsana and Asha Vijayan. Audio Steganography Using Modified LSB and PVD. Trends in Network and Communications. Communications in Computer and Information Science, 2011, Volume 197, Part 1, 11-20, DOI: 10.1007/978-3-642-22543-7_2.

[48] Shahidan M. Abdullah, Azizah A. Manaf, and Mazdak Zamani. Capacity and Quality Improvement in Reversible Image Watermarking Approach. 6th International Conference on Networked Computing and Advanced Information Management. 16-18 August 2010. Seoul, Korea.

[49] Shahidan M. Abdullah, Azizah A. Manaf, and Mazdak Zamani. Recursive Reversible Image Watermarking Using Enhancement of Difference Expansion Techniques. Journal of Information Security Research. Volume 1 Number 2. June 2010. Pages 64-70. PrintISSN: 0976-4143. Online ISSN: 0976-4151.

[50] Shih, F. Y. and Edupuganti, V. G. (2009). A Differential Evolution Based Algorithm for Breaking the Visual Steganalytic System. Soft Computing. Volume (13), 345-353.

[51] Shirali S., Manzuri-Shalmani M.T. (2007). Adaptive Wavelet Domain Audio Steganography with High Capacity and Low Error Rate. International Conference on Information and Emerging Technologies. 25-29.

[52] Sos S. A., David A. and Sunil A. (2005). Two Algorithms In Digital Audio Steganography Using Quantized Frequency Domain Embedding And Reversible Integer Transforms. Non-linear Signal Processing Lab. University of Texas. San Antonio, Texas, USA, 2005.

[53] Tewari, T.K., Saxena, V. & Gupta, J.P. 2011, "Audio watermarking: Current state of art and future objectives", International Journal of Digital Content Technology and its Applications, vol. 5, no. 7, pp. 306-314. International Journal of Digital Content Technology and its Applications. Volume 5, Issue 7, July 2011, Pages 306-314.

[54] Tiberiu M., Eric G., Mohamed N. (2002). Audio Digital Watermarking Based on Hybrid Spread Spectrum. Second International Conference on WEB Delivering of Music. 2002.

[55] Wang, R., Lin, C., Lin, J. (2001). Image Hiding By Optimal LSB Substitution And Genetic Algorithm. Pattern Recognition. Volume (34), 671-683.

[56] Wasan Shaker Awad and Huda Kadim Al Jobori. Non-Homogeneous Steganography Using Genetic Simulated Annealing. International Journal of Information Science and Computer Mathematics. Volume 2, Issue 1, Pages 61 - 73 (August 2010).

[57] Zanganeh, O. & Ibrahim, S. 2010, "Image steganography based on adaptive optimal embedding". 5th International Conference on Computer Sciences and Convergence Information Technology, ICCIT 2010. 2010, Article number 5711125, Pages 600-605.

[58] Mazdak Zamani, Azizah Bt Abdul Manaf, Shahidan M. Abdullah. Efficient Embedding for Audio Steganography. 2nd International Conference on Environment, Economics, Energy, Devices, Systems, Communications, Computers, Mathematics (EDSCM '13). Saint Malo & Mont Saint-Michel. France, April 2-4, 2012.

[59] Mazdak Zamani, Azizah Bt Abdul Manaf, Shahidan M. Abdullah, Saman Shojae Chaeikar. Correlation between PSNR and Bit per Sample Rate in Audio Steganography. 11th International Conference on Signal Processing (SIP '12). Saint Malo & Mont Saint-Michel. France, April 2-4, 2012.

[60] Mazdak Zamani, Azizah Bt Abdul Manaf, Shahidan M. Abdullah. Correlation between PSNR and Size Ratio in Audio Steganography. 11th International Conference on Telecommunications and Informatics (TELE-INFO '12). Saint Malo & Mont Saint-Michel. France, April 2-4, 2012.

[61] Shen Wang, Xianhua Song, Xiamu Niu, "An Affine Transformation Based Image Steganography Approach", JDCTA: International Journal of Digital Content Technology and its Applications, Vol. 6, No. 1, pp. 85 - 14, 2012.

[62] Jiaohua Qin, "An Improved Method to Secret Message Length Estimation in ±k Embedding Steganography", JCIT: Journal of Convergence Information Technology, Vol. 6, No. 5, pp. 339 - 346, 2011.

[63] Lingyun Xiang, Xingming Sun, Gang Luo, Bin Xia, "Steganalysis of Syntactic Transformation based Steganography", JDCTA: International Journal of Digital Content Technology and its Applications, Vol. 5, No. 5, pp. 320 - 330, 2011.

[64] Chiew Kang Leng, Jane Labadin, Sarah Flora Samson Juan, "Steganography: DCT Coefficients Reparation Technique in JPEG Image", JDCTA: International Journal of Digital Content Technology and its Applications, Vol. 2, No. 2, pp. 35 - 41, 2008.


122

An Overview on Audio Steganography Techniques Vol6 No13 Binder1... · 2012-07-31 · lower payload of the watermarking methods compared to steganographic algorithms [35, 42, 47, 63].

Documents