1

Hiding Encrypted Speech Using Steganography

Atef J. Al-Najjar, Aleem K. Alvi, Syed U. Idrees, Abdul-Rahman M. Al-Manea

Abstract: — In this paper, we have proposed a cryptosystem Steganography is used to hide the secret message data using
using state of the art technology. We used the minimum possible any other innocent cover object (may be text, image, audio or
compressed speech file for encryption. Additionally we applied video etc.) in such a way that another person could not know
data compression to speech files and added further compression.
Then we applied data encryption techniques to minimize security
about the existence of the original message. However it is
threats. We applied the symmetric and asymmetric encryption different with cryptography in the sense that cryptography
algorithms to the compressed speech files. This was followed by encrypts the data to be secured, which is generally understood
hiding these files into cover images using steganography. This and visible. Any one can try brute force or any other algorithm
camouflages the secret data and reduces chances of to break it.
eavesdropping. By using compression, the capacity has been
increased. However, after applying encryption, the size of the
encrypted data becomes much greater than the input file size, Capacity
therefore hiding it in a cover demands more capacity. We have
analyzed the size of the cipher texts of the compressed speech
data for further enhancement of the capacity.

Keywords — speech compression, symmetric and asymmetric

encryption, steganography

I. INTRODUCTION
Robustness Transparency
T his paper deals with the proposed cryptosystem including
compression, encryption and steganography. The Figure 1: Tradeoff between capacity, transparency and
proposed system incorporates these technologies in steps. The robustness [1]
objective of the system is to send the secret speech data with Steganography does not use algorithms to encrypt
security, transparency and robustness. Figure 1 shows the information. On the contrary, it is a process to hide data
relation between capacity, transparency and robustness. If we within another object so that it is hard to detect its presence. A
increase the capacity of any cover to store data with more than secret message can be hidden in a wav file, graphic file, exe
certain threshold value, then its transparency will be affected. file or html using tags etc.
With high capacity, the steganography is not strong enough to
keep transparent from the eavesdroppers. Similarly it is vise In this paper we hide the compressed speech data after
versa for robustness. It is a tradeoff between the capacity, applying selected cryptographic algorithms. Nowadays, secure
robustness and transparency. It is required to select all communication means the object data send by encryption
parameters in such a way that steganography can be achieved either it is half duplex or full duplex, real time or non real
on the best level. time, the security is one of the major parameters. This is the
war between eavesdroppers and cryptography experts who
Dr. Atef Jawad Al-Najjar is an Assistant Professor at the Department of develop the algorithms for secure communications. Always
Computer Engineering, College of Computer Science & Engineering, King the need of better secure communications in comparison of the
Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia, (contact at
e-mail: alnajjar@kfupm.edu.sa). existing ones is creating the motivations to develop more and
Aleem Khalid Alvi is a Lecturer–B and doing Ph.D. at the Department of more secure systems. However, by applying encryption
Computer Engineering, College of Computer Science & Engineering, King techniques, we can have encrypted data of our speech ready
Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia, (contact at
e-mail: akalvi@kfupm.edu.sa).
for transmission, but is it still secure? The answer depends on
Syed Usama Idrees is a Research Assistant and MS student at the the type of encryption techniques, the time required to break
Department of Computer Engineering, College of Computer Science & it. However, in addition to this, by using steganography
Engineering, King Fahd University of Petroleum & Minerals, Dhahran, Saudi
techniques we reduce the chances of suspicion on secret data
Arabia, (contact at e-mail: usama@kfupm.edu.sa)
Abdul-Rahman Muhammad Al-Manea is a EDP System Analyst at the from eavesdropper who caught the message object (cover).
SAP Computer Center, Saudi Aramco, and MS student at the Department of
Information & Computer Science, College of Computer Science & In symmetric key algorithms, when DES was broken by
Engineering, King Fahd University of Petroleum & Minerals, Dhahran, Saudi
Arabia, (contact at e-mail: mr.alman3@gmail.com) using special computing power [2], 3DES algorithm was
 2

suggested for increasing the key size three times that of DES internet, e-commerce, mobile telephones, banks, automatic
without changing the DES algorithm [3]. RSA as asymmetric teller machines, and in digital rights management, etc.
algorithm use longer keys for more security [4]. However, this
cannot eliminate the struggle against eavesdropping entirely. Cipher

Hidden messages can be concealed from eavesdroppers using

Cipher in terms of
steganography. It means he/she is unlikely to attempt an attack Cipher in terms of
Encryption process number of keys
on the captured data.

II. COMPRESSION Block Stream Asymmetric Symmetric

cipher key cipher key cipher
cipher (Secret key
Preparation for the encryption step includes the use of the (Public key
ciphers) ciphers)

smallest size secret speech data using the appropriate file

format without loosing the quality of the secret speech. We Symmetric Asymmetric
key cipher key cipher
have taken one speech sample that includes three 'Hello' word (Secret key (Public key Block Stream
block ciphers) ciphers) cipher cipher
voice. We store this sample using 7 uncompressed, 6 lossy
compressed and 1 lossless compressed formats. For further
compression we apply the LZ77 algorithm [5, 6] using zip Figure 3: A taxonomy of cryptosystems based on [7]
utility to compress. We have found little improvement in size
with certain audio formats. Figure 2 shows the comparison
among these audio codec algorithms. Figure 3 shows the taxonomy of the cryptosystems in term
We have observed that there is no change in size for the of the number of keys and encryption process. We have used
uncompressed formats and also for some compression the secret key cipher and public key cipher i.e. 3DES (cipher
formats. In lossy-compression, we have observed that there is block chaining), Asymmetric Encryption Standard (AES), and
much compression. However, after the LZ77 algorithm was RSA, respectively.
applied, some of them showed improvement and some did not.
In lossless compression, there is improvement in decreasing Encryption creates the secret encrypted message for
size, but no improvement using the LZ77 algorithm. protecting the confidentiality. However for integrity,
authorization and authenticity we need further techniques; for
example, a message authentication code (MAC) or digital
Comparison of different audio formats
(with and with out zip compression) signatures. Cryptosystems are based on cryptographic
Audio samples with out zip compression
200 Audio samples with zip compression
software and hardware to perform encryption and decryption.
Size in KB

150 Uncompressed file format However, successfully using cryptosystems to ensure security
100 Lossless compression format is not hundred percent sure. Any weakness or flaw in the
Lossy compression format
50 system design or execution can allow successful attacks from
0 the eavesdroppers.
av

w
u

.rm

a
64

ox
ca
ov

if
ig

p3
gg
.iv

m
.a

.a
.ra
.d

.w

.v
.p
.m

.w

.m
.o

.w

IV. STEGANOGRAPHY
Audio codecs extensions
Today, many algorithms show their stability for secure
Figure 2: A comparison of audio samples communications, e.g. RSA with large key size, say 1024, as
public key cipher or AES as private key cipher. However it
We found that the Windows Media Audio (WMA) format provides more security if eavesdroppers can not detect the
is the best candidate among others for encryption and then presence of the secret message either encrypted or not. If
steganography. Because WMA file format gives the smallest encrypted secret message is detected then eavesdropper starts
file size (with using specific settings for encoding). We found to work out the message by applying the technique to decrypt.
a further reduction of 1KB in size from the original (WMA At this stage cipher algorithms must be sophisticated that the
file of sample) after applying the LZ77 algorithm. time for breaking it, is must be beyond the present context.
An encrypted hidden message into a cover object makes
eavesdropping more difficult.
III. ENCRYPTION
Encryption is the process of transforming information; e.g. The word “Steganography” is of Greek origin and means
plaintext, into encrypted text called ciphertext. That ciphertext “covered or hidden writing”. Hiding media may be text,
can only be decrypted by having specific information. image, audio, video, binary or html file etc. This apparent
Decryption refers to the reverse process. media object looks as a message called cover. For instance, in
Encryption/decryption has been used from ancient times to old ages, a message may be hidden by using invisible ink
modern age for secret communication. Nowadays, use of between the visible lines of innocuous documents. Modern
security is for communication either real time or not, in steganographic techniques differ with the historical ones.
 3

There implementation require algorithmic procedures and are V. METHODOLOGY

complex in nature. However, it provides more security. Figure We have taken advantage of both encryption and
4 shows the classification of the information hiding presented steganography in our system, for enhancing security. We
in [8, 9, and 10]. Based on this taxonomy we use modern have selected the speech file as the secret message stored into
steganography with images for hiding the encrypted data. a smallest file format (e.g. wma). For further compression we
apply the LZW compression algorithm and store it into zip file
Steganography has the advantage to cryptography that format. This finally compressed file is encoded into integer
secret messages do not attract attention to an eavesdropper. If numbers using MATLAB.
the message is encrypted but not hidden; it will arouse
suspicion and may be in itself be the victim of the attacker, We use the selected RSA from asymmetric category and
whether he or she breaks it or not, but attempts are done. This 3DES and AES (Rijndael) from symmetric encryption
also makes incriminating on sender when encryption is illegal algorithms to encrypt the integer numbers. The selection is
in their countries [11]. made because of there diverse algorithmic procedure.
However; the 3DES is the ancestor of AES. This encrypted
In some application scenarios, multimedia steganography data is hidden in a cover object (we use the image) using a
techniques have to satisfy two basic requirements in order to simple steganography technique (2 bit LSB insertion with
make the cover object and stego object perceptually minimal affect on the perception of the cover object.). Then,
indiscernible. The first requirement is perceptual transparency the given cover has been modified to store the encrypted
and the second one is high data rate of hidden data. With few compressed speech data. The resulting cover object is a stego
changes, every application requires a high bit rate of an object without showing significant change. The whole
embedded data [12]. sending-end process of the cryptosystem is shown in Figure 5.
COVER
A cover object (image, audio or video etc) with a large size OBJECT

(means comprising of a large number of bits) can be used to 21 34 67 79 36 10 1010110

10100101
01101001
hide the secret message with more ease. Because it has good 91 87 39
-------
19 71 01
-------
0110011
-------
00101001
01101010
------- ------- ------- - - - - - - -- -
capacity to store the data without revealing any change in the SECRET
SPEECH ------- ------- ------- ---------
------- ------- ------- ---------
original cover object. For this reason, a digital cover object FILE
------- ------- ------- ---------
31 47 57 11 91 73 1000110 ---------
containing large amounts of data is used to hide messages on 11010100

the internet and on other communication media. Apply LZW compression Open zip file in Encryption Binary conversion Hide data using
integer numbers steganography

A cover can be created in any way, e.g. 2 least significant Figure 5: Sending-End of the proposed cryptosystem
bits of each byte of an image can be replaced with the bits of Now only the recipient knows the technique to recover the
the secret message. This LSB replacement does not greatly message and then decrypt it using the receiving-end part of the
affect the cover object (image) to reveal the change. cryptosystem, discussed later

The other practice of the information-hiding is “A digital A. Sending End

watermark”; that can be inserted on the digital images so that We have been selecting speech file format with least size by
illegal copies of the images can be detected with some more experimenting with available codec’s. The best we have found
sophisticated method. This practice is also carried out while is the WMA format, which is smaller in size and preserving
printing currency notes at the monetary agencies. the quality of the speech. The smaller size leads to faster
computation during the encryption/decryption process, and
Information reduces the power consumption (if battery is used) of the
Hiding
overall system in case of handheld devices.
Covert channels Steganography Anonymity Copyright marking
The purpose of hiding encrypted compressed speech files
into cover object (using steganography), is to mislead the
Linguistic Technical Robust copyright Fragile
marking watermarking eavesdropper. This will minimize the threat of eavesdropping.
Covered cipher Audio
Encrypted compressed speech file inside image as cover
Fingerprinting Watermarking object is a unique combination of hiding the message to
Open cipher Image
achieve more secure communication. For increasing the
Misspellings
Semagrams Text Imperceptible Visible capacity, audio or video files can be used for the cover object.
watermarking watermarking
However, using an image as the cover object, capacity has
Phonetics
Audio been increased by selecting a least size gray image or three
Jargon Image
dimensional images etc.

Figure 4: A classification of information hiding techniques

based on [8, 9, 10]
 4

Table 1: Comparative analysis of data compression for 6

speech . It receives the stego object and retrieves the hidden
encrypted data. This encrypted data is converted from binary
Comparative Analysis of Data Compression for
Audio / Speech to integer numbers and then decrypted using the decryption
Types Size in Kilo Bytes algorithm. This decrypted data is decoded into the zip file and
Samples file formats wma only decompressed using the LZW compression algorithm (using
wav wma zip rar zip utility). The result is the secret speech file.
Hello (3
General 173 6 5 5
times)
COVER
Bismillah OBJECT
940 8 6 6
(Arabic) 10100101
01101001 1010110 79 36 10 21 34 67
Bismillah 00101001
01101010
0110011
-------
19 71 01
-------
91 87 39
-------
(English 747 6 5 5 - - - - - - -- - ------- ------- ------- SECRET
SPEECH
--------- ------- ------- -------
Translation) --------- ------- ------- ------- FILE
--------- ------- ------- -------
Quranic Bismillah --------- 1000110 11 91 73 31 47 57
11010100
Verses in (Urdu 809 7 5 5
Arabic and Translation)
their Surah-e- Extraction of binary secret
data from cover object
Conversion from Decryption Integer numbers
Convert into zip file
Extract secret speech file using
LZW decompression
binary to integer
translations Ikhlas 2463 16 13 13
in English (Arabic)
and Urdu Surah-e- Figure 6: Receiving-End of the proposed cryptosystem
languages Ikhlas
2465 17 14 14
(English
Translation)
We used a number of samples of speech, in different
languages (i.e. English, Arabic, Urdu), to get a mix of
Surah-e-
Ikhlas (Urdu 3048 19 16 16 frequencies. This was done to enhance the accuracy of the
Translation) results. Therefore a reverse procedure has the following steps,
i.e. (1) the removal of the bits from the stego, (2) conversion
into integers, (3) decrypting the temporary result with the
Selecting the least file size using WMA codec and then decryption key and (4) finally decompressing; that gives the
applying a compression algorithm to reduce the data size original secret speech. This whole process is shown in Figure
yields good results (in terms of decreasing the size). Table 1 6.
shows the comparative results of data compression. The Cipher text comparision
samples text is mention in appendix A.
500 RSA

Table 1 describes the significant change in size from wav to 400 AES
Size in KB

using WMA (codec) with maintaining the quality of speech. 300 3DES
Additionally, by applying the LZW compression algorithm, 200
further reduction in size is obtained. However there is no 100
difference between zip and rar compression. 0
1 2 3 4 5 6 7
We have done additional experiments to understand the Samples
effect of the size of the secret message in terms of the
different linguistics. We have selected three languages: Figure 7: A comparison of cipher text
Arabic, English and Urdu for this purpose. We have taken the
same secret message in all three languages recorded in wav Figure 7 show that the comparison of the cipher text after
and WNA formats. It is observed that different file sizes were encryption and that gives the requirement of the capacity of
obtained for the same message in different languages, and the cover object. This shows that the RSA (asymmetric
Urdu has taken more file size in comparison of Arabic and algorithm) has more cipher text in size as compared to the
English. We left it for future research to repeat the same 3DES and AES (symmetric algorithms). In this case, the less
experiment with as many languages as possible, and find the capacity symmetric algorithm is best to encrypt the data. On
language in which the secret message is always stored the other hand, if the capacity is available, (e.g. cover object
as video) then RSA will be preferable because it solves the
(statistically) in a lower size. Then we can manage the
key management problem.
language translation for from one end of the crypto system to
another.
In [13], Lisa M. Marvel elaborates on the capacity of a
B. Receiving End cover by developing a capacity metric that can be used as a
The receiving-end of the cryptosystem performs the reverse performance criterion for the class of steganographic methods
that add the secret speech data. It is used for all types of cover
of all steps which have been done before. The detailed
data including imagery and audio. It is used to calculate
diagram of the receiving-end cryptosystem is shown in Figure
precise estimate of the maximum amount of payload that can
 5

be embedded and successfully extracted from the cover data. [13] Lisa M. Marvel, Image Steganography for Hidden Communication,
Ph.D. Dissertation, Spring 1999, University of Delaware, US.Kaushal
[14] M. Solanki, Multimedia Data Hiding: From Fundamental Issues to
In [14] Kaushal M. Solanki have introduced a general Practical Techniques, Ph.D. Dissertation, December 2005, University of
framework for a data hiding system, based on the following California, Santa Barbara, US.
steps i.e. (1)encoding of message & host image with secret
key, (2) generation of a composite image, (3) avoiding VII. APPENDIX A
intentional/unintentional processing attacks, (4) decoding with In this appendix A we have given he scripts of the all the
the secret key to get original message. speeches which we have used in the Table 1.

VI. CONCLUSION Bismillah:

Arabic
We have discussed vital issues of a cover object that
contains essential speech data by saving compressed speech
files in different languages using lossless and lossy audio English translation:
formats. We have shown the results for the overall operations.
As a complete cryptosystem, data encryption techniques In the name of Allah, Most gracious, most merciful
applied on compressed speech minimizes security threats.
Steganography was further used to hide the encrypted data for
improved security. It minimizes the threat of eavesdropping. Urdu translation:

The cipher texts of the compressed speech data shows that

the asymmetric algorithm produces large cipher text file in
comparison of the symmetric algorithm. Figure 7 shows that Surah-e-Ikhlas:
in case of utilization of asymmetric algorithm always we need Arabic
approximately 5.7 times more capacity in cover object as
compared to symmetric algorithms.

REFERENCES
[1] J. Fridrich, Applications of data hiding in digital images, Tutorial for the
ISSPA 1999 conference in Brisbane, Australia, 1999.
[2] Fred Moore, "Preparing for encryption: new threats, legal requirements
English translation1:
boost need for encrypted data", Computer Technology Review, August-
Sept, 2005
[3] NIST, "Recommendation for the Triple Data Encryption Algorithm In the name of God, Most Gracious, Most Merciful
(TDEA) Block Cipher" , Special Publication 800-67. [112:1] Proclaim, "He is the One and only GOD.
[4] R. L. Rivest and R. D. Silverman, `Are "strong" primes needed for [112:2] "The Absolute GOD.
RSA?', Preprint , 1999, 1--23. http://citeseer.ist.psu.edu/rivest99are.html
[112:3] "Never did He beget. Nor was He begotten.
[5] Jacob Ziv and Abraham Lempel; A Universal Algorithm for Sequential
Data Compression, IEEE Transactions on Information Theory, Vol.23, [112:4] "None equals Him."
No.3, pp.337-343, May 1977.
[6] Vito Dai and Avideh Zakhor, Lossless Compression Techniques for
Urdu translation2:
Maskless, Video and Image Processing Lab, Department of Electrical
Engineering and Computer, Univ. of California/Berkeley, http://www-
video.eecs.berkeley.edu/papers/vdai/spie4688-67.pdf
[7] Ian Barnes, Ramesh S., Lecture on Public Key Cryptosystem,
Department of Computer Science, Faculty of Engineering and
Information Technology (FEIT) Fall semester 2006, The Australian
National University , Accessed on June 29, 2007,
http://cs.anu.edu.au/Student/comp3410/lectures.html
[8] F. L. Bauer, Decrypted Secrets-Methods and Maxims of Cryptology,
Berlin, Heidelberg, Germany: Springer-Verlag, 1997.
[9] Fabien A. P. Petitcolas, Ross J. Anderson and Markus G. Kuhn,
Information Hiding - A Survey, Proceedings of the IEEE, special issue
on protection of multimedia content, Vol.87,No.7, pp.1062-1078, July
1999.
[10] D. Vitaliev, Digital Security and Privacy for Human Rights Defenders,
The International Foundation for Human Right Defenders, pp.77-81,
Feb. 2007.Lisa M. Marvel, Image Steganography for Hidden
Communication, Ph.D. Dissertation, Spring 1999, University of
Delaware, US.
[11] Caloyannides, M.A., Encryption wars: shifting tactics, Spectrum, IEEE,
Vol. 37, No. 5, pp.46 – 51, May 2000
[12] R.J. Anderson, F. A. P. Petitcolas, On the Limits of the Steganography,
1
IEEE Journal of Selected Areas in Communications, Vol. 16, No. 4, Islamic portal, http://www.quran-islam.org/110.html
2
1998, pp. 474-481. Comprehensive portal of Urdu
http://www.loveurdu.com/quranurdu/home.asp?Pid=547