Professional Documents
Culture Documents
Adaptive Audio Steganography Scheme Based On Wavelet Packet Energy
Adaptive Audio Steganography Scheme Based On Wavelet Packet Energy
Abstract—This paper proposes an adaptive audio steganog- for 83% and 11%, respectively. It can be seen that auditory
raphy scheme based on wavelet packet energy which can system is the largest source of information in addition to
be used to hide secret messages in digital audio. The major the Human Visual System(HVS). Thus, the research of
contribution of the proposed scheme is that the wavelet packet
sub-bands of the host audio can be selected adaptively to embed audio steganography is of great significance and has wild
the secret messages according to the masking effect of human application scenarios. Especially in the era of big data, audio
auditory system (HAS) and weighted energy concentration . steganography will have a more broad prospects in copyright
First, the host audio is divided into many segments which are protection, secret information transmission, covert channel
then decomposed by wavelet packet to calculate the wavelet and etc.
packet energy feature. Second, the covert data are embedded
into each segment through adjusting the relationship among Like steganography in other media, audio steganographic
wavelet packet sub-bands by modifying or exchanging the co- technique also has three evaluation metrics [3]:
efficients of wavelet packet sub-bands. Finally, the experimental
results and comparison with existing technique show that the
1) Capacity means the amount of secret information that
proposed scheme has larger hiding capacity while maintaining can be embedded into the host audio without affecting
imperceptibility and strong robustness. the perceptual quality of audio.
Keywords-Audio steganography; Wavelet packet; Weighted 2) Imperceptibility evaluates how well a secret message
energy concentration; Adaptive; Hiding capacity. is embedded into the cover audio. The difference
between audio after hiding and audio before hiding
I. I NTRODUCTION should remain negligible.
3) Robustness indicates the ability of secret messages to
Steganography is an art of hiding secret information in
resist against attacks.
another seemingly innocuous message, or carrier [1]. Com-
pared with encryption technique [2], steganography conceals In audio steganography, HAS is used to hide informa-
the existence of secret messages instead of readability. Since tion in the audio. Because the human auditory system has
when an attacker doesn’t know the existence of secret more precision than HVS, audio steganography has more
messages, he won’t generate the idea of attacking it. So, challenges than image steganography [4].
steganography is a powerful technique which enhances se- In this paper, we propose a scheme that has high capacity
curity in data transferring and archiving. Throughout history, and good output quality. The proposed scheme is based on
a multitude of methods have been used to hide informa- wavelet packet transform with adaptive hiding in wavelet
tion. With the development of the Internet and other new packet sub-bands. The adaptive hiding is determined by
technologies, digital steganography technique which is used the sub-band weighted energy concentration. The messages
to embed the secret messages into digital multimedia is can be recovered without original host audio. Experimental
gradually rising. It has developed a strong basis for the area results and comparison with existing method show that the
of steganography with a growing number of applications proposed scheme has an advantage of large hiding capacity
for digital fields like digital rights management, covert while maintaining imperceptibility and strong robustness.
communications, annotation etc. So far, various researches The remainder of this paper is organized as follows:
on steganography have been carried out on storage media, Section II discusses some existing methods. The proposed
such as text, image, audio, and video. scheme is introduced in section III. In section IV experi-
According to the education experts, human beings receive mental results of proposed scheme are discussed. Finally,
information mainly by visual and auditory, which accounts Section V concludes this paper.
27
sequence. Then the weighted energy concentration is calcu-
Host Audio Segmenting
Wavelet Packet lated. The formula as follows:
Decomposition
q
Energy (16 − p) · SP
Calculating Sub- Weighted Energy Threshold
p=0
band Energy Concentration wE = , q = 0, 1, . . . , 15 (2)
15
Embedding Sub-
(16 − i) · Si
band Selection i=0
28
Table I
E MBEDDING RULES OF THE P ROPOSED S CHEME
Host Audio Wavelet Packet
Segmenting
Decomposition
Relationship of Sub-band Covert Relationship of Sub-band Covert
Energy Data Energy Data
(4,a)>(4,b)>(4,c)>(4,d) 0000 (4,c)>(4,a)>(4,b)>(4,d) 1000
Energy
(4,a)>(4,b)>(4,d)>(4,c) 0001 (4,c)>(4,a)>(4,d)>(4,b) 1001 Calculating Sub- Weighted Energy Threshold
band Energy Concentration
(4,a)>(4,c)>(4,d)>(4,b) 0010 (4,c)>(4,b)>(4,d)>(4,a) 1010
(4,a)>(4,d)>(4,b)>(4,c) 0011 (4,c)>(4,d)>(4,b)>(4,a) 1011 Embedding Sub-
band Selection
(4,b)>(4,a)>(4,c)>(4,d) 0100 (4,d)>(4,a)>(4,c)>(4,b) 1100
(4,b)>(4,c)>(4,d)>(4,a) 0101 (4,d)>(4,c)>(4,a)>(4,b) 1101
(4,b)>(4,d)>(4,c)>(4,a) 0110 (4,d)>(4,c)>(4,b)>(4,a) 1110 Camparing the
Ensuring the
First and Second
(4,b)>(4,d)>(4,a)>(4,c) 0111 (4,d)>(4,b)>(4,c)>(4,a) 1111 Relationship
Half Energy
Recovering the
Secret Message
ni /2
Secret
j=1
Algorithm 1. The embedding procedure of the proposed scheme IV. E XPERIMENTAL R ESULT AND A NALYSIS
Input: covert data and host audio
Output: stego audio In order to test the properties of the proposed scheme, we
Step 1: Segment the host audio, then perform integer wavelet packet design some experiments as follows. Quantitative as well as
decomposition.
Step 2: Calculate weighted energy concentration and sub-band energy,
qualitative analysis are performed for the proposed scheme.
then compare with energy threshold, and select embedding position. We tested with mono host audio belonging to category of
Step 3: According the covert data, adjust the relationship of the selected audio file named Pop Music and a text file named The
sub-bands. Then calculate the first and second half energy and according
the covert data, adjust the first and second half coefficients.
Little Prince (insert). The pop music is sampled at 44.1
Step 4: Perform inverse wavelet packet decomposition and after KHz with length of about 150 seconds and quantized by
recombining the fragments, get stego audio. 16 bits. It should be noted that other audio files as host
29
Figure 4. Original Host Audio Figure 5. Stego Audio in the Condition of No Attack
Table II
audio can also get the similar results. The proposed scheme T EST R ESULTS OF THE P ROPOSED S CHEME
is implemented using MATLAB (2013a) programming. Each
Scheme proposed Scheme proposed
segment contains 512 sampling points and threshold is set to Attacks in this paper in [21]
0.90. In this section, we utilize BER (Bit Error Rate), SNR BER(%) SNR(db) BER(%) SNR(db)
(Signal to Noise Ratio) and PESQ (Perceptual Evaluation 30 db SNR white noise 3.245 9.854 0.018 12.702
of Speech Quality) to evaluate the performance. Assessment 40 db SNR white noise 0.0168 13.242 0.015 13.857
of the quality of stego audio is made through P.862 (PESQ) re-quantization
0 ∞ 0.013 13.856
(16 bits∼32bits∼16bits)
which is recommended by ITU and in the range [1,4.5]. If re-quantization
0.0423 12.551 7.220 9.624
PESQ>3.5, the quality of audio accord with the standard of (16 bits∼8bits∼16bits)
telephone. Three formulas for calculating BER and SNR are up-sampling 0.0124 14.045 0.013 13.856
down-sampling 0.0618 12.137 2.810 11.632
given as follows,respectively.
30
be embedded into one segment. Compared with the scheme [7] Vimal, J., Alex, A.M.: Audio steganography using dual
proposed in [21], 8 bits can be embedded into one segment randomness lsb method. In: International Conference on
Control, Instrumentation, Communication and Computational
in proposed scheme, which is four times larger than 2 bits.
Technologies. (2014)
Provided that the length of one segment is 20ms, the capacity
of former is 100 bps while the latter is 400 bps. [8] Chen, T.C., Wu, W.C.: Highly robust, secure, and perceptual-
quality echo hiding scheme. Audio Speech & Language
V. C ONCLUSION Processing IEEE Transactions on 16(3) (2008) 629–638
This paper proposed a novel scheme which can adaptively
[9] Yan, B., Sun, S.H., Lu, Z.M.: Improved echo hiding using
select wavelet packet sub-bands to embed covert data into power cepstrum and simulated annealing based synchroniza-
digital audio utilizing wavelet packet decomposition and tion technique. (2003)
weighted energy concentration. The scheme makes it pos-
sible to embed covert data somewhere with satisfying the [10] Rupanshi, Preeti, Vandana: Audio steganography by direct
sequence spread spectrum. International Journal of Computer
weighted energy concentration based on auditory masking
Trends & Technology 13(2) (2014)
effect of HAS. Experiments show better properties than
existing methods. The quality of host audio is less affected [11] Chen, M., Zhang, R., Liu, F.F., Niu, X.X., Yang, Y.X.: Audio
as the embedding position of covert data is in middle or large steganography by quantization index modulation in the dct
frequency sub-band. Moreover, it is evident that the scheme domain. Journal on Communications 30(8) (2009) 105–111
not only has large capacity which is 4 times lager than that of [12] Premalatha, P., Narayanan, K., Vikash, Ramesh: Steganogra-
the method proposed in [21] , but also has strong robustness phy in audio signals using variable bit replacement method
for resisting the common attacking behaviors, such as noise, in dct domain. Esrsa Publications (2014)
re-quantitative. In addition, the greater the number of audio
[13] Chilhate, K., Patidar, K., Chandel, G.S.: Advanced audio
segments is, the larger the total hiding capacity is. Extraction steganography technique based on coefficient comparison in
process doesn’t need the original host audio to achieve the dct domain. (2015)
blind extraction. The future work will pay more attention to
implement lossless extraction of secret messages. [14] Sheikhan, M., Asadollahi, K., Shahnazi, R.: Improvement of
embedding capacity and quality of dwt-based audio steganog-
ACKNOWLEDGEMENT raphy systems. World Applied Sciences Journal 13(3) (2011)
This work is supported by the National Natural Science [15] Bhattacharyya, S., Sanyal, G.: A robust image steganography
Foundation of China (Grant Number: 61602491) and Hefei using dwt difference modulation (dwtdm). International
Electronic Engineering Institute Foundation (KY16A654). Journal of Computer Network & Information Security 4(7)
(2012)
R EFERENCES
[16] Cvejic, N., Seppanen, T.: Increasing the capacity of lsb-based
[1] Johnson, N.F., Duric, Z., Jajodia, S.: Information Hiding:
audio steganography. In: Multimedia Signal Processing, 2002
Steganography and Watermarking-Attacks and Countermea-
IEEE Workshop on. (2003) 336–338
sures. Springer Publishing Company, Incorporated (2012)
[17] TVS, G.P., Varadarajan, S.: A novel hybrid audio steganog-
[2] Gai, K., Qiu, M., Zhao, H., Xiong, J.: Privacy-aware adaptive
raphy for imperceptible data hiding. In: 2015 International
data encryption strategy of big data in cloud computing. In:
Conference on Communications and Signal Processing (ICC-
IEEE International Conference on Cyber Security and Cloud
SP). (April 2015) 0634–0638
Computing. (2016) 273–278
[18] Nehete, S., Sawarkar, S.D., Sohani, M.: Digital audio
[3] Divya, S.S., Reddy, M.R.M.: Hiding text in audio using
steganography using dwt with reduced embedding error and
multiple lsb steganography and provide security using cryp-
better extraction compared to dct. In: Icwet ’11 International
tography
Conference & Workshop on Emerging Trends in Technology,
[4] Shirali-Shahreza, S., Manzuri-Shalmani, M.T.: Adaptive Mumbai, Maharashtra, India, February. (2011) 167–168
wavelet domain audio steganography with high capacity and
low error rate. In: International Conference on Information [19] Shah, P., Choudhari, P., Sivaraman, S.: Adaptive wavelet
and Emerging Technologies. (2007) 1 – 5 packet based audio steganography using data history. In:
IEEE Region 10 and the Third International Conference on
[5] Cvejic, N., Seppanen, T.: Increasing robustness of lsb audio Industrial and Information Systems. (2008) 1–5
steganography using a novel embedding method. In: Inter-
national Conference on Information Technology: Coding and [20] Shahadi, H.I., Jidin, R.: High capacity and inaudibility
Computing, 2004. Proceedings. Itcc. (2004) 533–537 Vol.2 audio steganography scheme. In: International Conference
on Information Assurance and Security, Ias 2011, Melacca,
[6] Roy, S., Parida, J., Singh, A.K., Sairam, A.S.: Audio Malaysia, December. (2011) 104–109
steganography using lsb encoding technique with increased
capacity and bit error rate optimization. In: International [21] Qiuling, W., Meng, W.: Adaptive digital audio steganography
Conference on Computational Science, Engineering and In- based on integer wavelet transform in chinese. Journal of
formation Technology. (2012) 372–376 Electronics & Information Technology 38(4) (2016) 834–840
31