ISSN 2347 - 3983

Volume 7, No. 9 September 2019

Reggie C. Gustilo International Journal
et al., International Journal of Emerging
of Emerging Trends
Trends in Engineering in7(9),
Research, Engineering
September 2019,Research
346 - 352
Available Online at http://www.warse.org/IJETER/static/pdf/file/ijeter19792019.pdf
https://doi.org/10.30534/ijeter/2019/19792019

Android-based Image and Video Steganography System

Reggie C. Gustilo1. Ron Michael Castillo2, Nicole Anthonnie Gonzales3, John Gerard Raz4, Trisha Jane Tejones5
Electronics & Communications Engineering Department
De La Salle University Manila, Philippines
1
reggie.gustilo@dlsu.edu.ph
2
ron_castillo@dlsu.edu.ph
3
nicole_gonzales@dlsu.edu.ph
4
john_gerard_raz@dlsu.edu.ph
5
trisha_tejones@dlsu.edu.ph

ABSTRACT assuring that the information is accurate. There are few

instances that information must be properly secured or even
Steganography is a form of security measure wherein confidential concealed to protect it from unauthorized access from
information is hidden within a cover media. This study aims to unknown users [19].
contribute to the security of digital information, particularly of
audio data using image and video steganography. A total of six
technique combinations were derived from three-pixel location Cryptography protects the information being sent by
strategies and spatial domain embedding in order to successfully converting it into data that is hidden or unreadable to an
mask and impose audio data into cover images and videos. The unknown user. This is done to secure digital information sent
final prototype comes in the form of an integrated system that over unbounded or unsecure media. A key or password is
operates on three environments. The first platform is an android usually shared by the sender and receiver to extract and
application, which serves as the input/output gateway, followed recover the information signal. Cryptography provides
by a Matlab program which serves as the main processor, and a confidentiality, authenticity and data integrity for computer
server that allows two-way exchange of information at both ends. systems. However, because of the advancements in
The goal of the prototype is to successfully hide the audio data into
technology and computer science, cryptography based
the cover media with minimal to negligible visual and audible
changes between the original files and output files. To evaluate security schemes are vulnerable, at the same time
the system’s performance, both qualitative and quantitative bases cryptographic techniques are hard to decode. While messages
are used. Quantitative assessment includes the MSE and PSNR of are written in secret codes and symbols, it is apparent to other
the steganography output, the BER and accuracy of the recovered users that it is an encrypted message. The contents of the
audio data. Meanwhile, qualitative assessment is conducted in the messages are hidden while the message itself is not.
form of a survey. Results from quantitative and qualitative
assessment strongly verify that the proposed integrated system is A new technology called digital steganography is an
verified and hereby recognized as consistent, reliable and alternative to cryptography. Unlike cryptography,
accurate.
steganography hides both the content and any signs of
hidden message from information hackers. Digital
Keywords: Steganography, image, video, embedding
steganography embeds an encrypted message in a cover
media such as audio, digital image or video. It is a technique
1. INTRODUCTION
wherein the least significant components of a cover media
are replaced by the hidden message. With this technique,
In this digital age, hundreds of researchers focus on topics other users are kept unaware of the existence of the message.
that take advantage of the advancements of technology and The steganography image or video should be visually similar
computer systems. Some of these research topics cover to the original file to avoid suspicions of having hidden
machine vision system [1], [2], [3], [4] and expert systems or messages.
intelligent systems [5], [6], [7], [8], [9], [10]. [11], [12]. On
the other hand, all these research topics must assure that all Image Steganography [20] usually uses the Least Significant
information are secure and accurate even if this information Bit technique in embedding the message on the images.
are transmitted in public domains or unbounded media like Images typically has either 8-bit or 24-bit color matrix. The
wireless transmission such as wifi, Bluetooth and the like. pixels are composed of a set of bits that define its color. The
The information being transmitted can be altered resulting to encrypted message or the cipher-text are embedded in the
errors during transmission. Fortunately, there are also lots of message in such a way that alterations are not distinguishable
researches that focus on how transmission of data can be by other users. The Least Significant Bit technique alters the
improved [13], [14], [15], [16], [17], [18]. Aside from
346
Reggie C. Gustilo et al., International Journal of Emerging Trends in Engineering Research, 7(9), September 2019, 346 - 352

LSB of a pixel that contains the bits of the hidden message. 2. PROBLEM STATEMENT
The LSB method is the most common and the simplest
technique used for embedding messages in a cover media. A complex method of hiding information using audio signals
The LSB method also allows embedding large capacity of was proposed using image and video steganography. Audio
information [4] signal considered information signals are embedded in either
an image cover media or a video cover media.
Video Steganography uses video files as cover media. The
use of video steganography is one of the most secure type of 3. METHODOLOGY AND DESIGN
steganography due to its capacity and complexity. The video CONSIDERATIONS
file is converted into frames where the bits of the encrypted Here are the techniques and methods used in this research to
message are embedded. The use of the video file as the cover fully implement an audio and video steganography.
media, the bits of the message or information embedded is
unnoticeable.

Table 1: Elements of the Image and Video Steganography

Software/ Method
Description
used
This is used to perform most of the processing like embedding of the audio data into the cover
medium, extracting audio data from previously embedded files, conversion and sampling of
Matlab
audio data, password generation, recovery and analysis, and technique-based identification of
pixel locations.
This is used develop an Android Application that will serve as the input and output interface
Android Studio
of the project.
It acts as the medium between the Android Application and Matlab. Its main purpose is to
Server
store data for both the android application and Matlab.
The Android Application developed is capable of reading audio, image, and video files from
Android Application internal storage. It is used for uploading and downloading audio, image, and video files to
database server.
Audio as a Secret
Audio files were used as the secret messages in this study.
Message
Image as a Cover Image files are used to embed audio files to cover the hidden information. The last bit of each
Media color component of an image is used as storage for the audio message to be transmitted.
Images from video files are used to embed audio files to cover the hidden information and can
Video as a Cover
carry larger audio data than image cover media. Mp4 format is not used due to compression
Media
done that corrupts the embedded audio message.
The LSBs of the selected pixels are compared and modified in accordance to the secret
LSB Embedding message. The R, G and B components of each pixel serve as carriers of the secret message, by
Technique modifying the last 2 or 3 bits of the color values. For a single pixel, a maximum of 9 LSBs
may be modified.
Selection of Pixel These are methods that choose the location of the pixels that will carry the information signal
Location Strategies or the audio signal.
Password is used increase the security of the information. It consists of 10 alphanumeric and
Password special characters that are generated in both image and video embedding. Passwords have a
total length of 57 bits.

3.1 Embedding Technique Design T4 Pixel Intensity Based Technique 3

T5 Side-most Technique 2
Six techniques were derived and implemented from the T6 Side-most Technique 3
combinations of three carrier pixel location strategies and
spatial- domain embedding. Table 2 below shows the bases 3.1.1 Pseudo-Random Pixel Location Selection
of each technique in terms of carrier pixel location and n-bit Depending on the dimensions of the image, this technique
embedding that needed to hide the information. produces a specific sequence of the designated carriers and
all the pixels contained in the image can be used. The
Table 2: Technique Details technique will first identify the locations of each pixel then it
Technique Pixel Location Strategy LSB will rearrange the said locations using the sequence
T1 Pseudo Random Technique 2 produced by the technique.
T2 Pseudo random Technique 3
T3 Pixel Intensity Based Technique 2

347
Reggie C. Gustilo et al., International Journal of Emerging Trends in Engineering Research, 7(9), September 2019, 346 - 352

3.1.2 Pixel Intensity Based Selection 3.2 Password Generation

In this technique, the pixels recognized as dark are For security measures, passwords consisting of 10
designated as the carriers of the audio data. The grayscale alphanumeric and special characters are generated in both
values of the pixel images determine the darkness or image and video embedding. The password’s binary data
whiteness of the pixel. The dark pixels are then identified comprises of a total of 57 bits which corresponds to 7
and used as carriers. pixels (three color value RGB for each pixel, 3 bits per
color value). The sampling frequency of the audio
In the identifying the carrier pixel locations, the equivalent information used 18 bits of data that requires to pixels for
grayscale image of the original true colored image aree used embedding and lastly, the number of channels of the
as the basis for visual examination. The equation for getting audio information that requires 2 bits and embedded in a
the equivalent gray values are shown below. red color data of a pixel. Table 3 below shows the bit
configuration of the password for both image and video
Gray value = 0.299R + 0.587G + 0.114B (1) cover media.

Where R = red pixel component value Table 3: Bit Configuration for Passwords
G = green pixel component value Image Video
Nth
B = blue pixel component value password password
Character
bit count bit count
Pixels with a grayscale values ranging from 0 to 120 are 1 7 3
assigned as carrier pixels. The maximum absolute deviation 2 3 7
from the original pixel value is 7 due to the fact that up to 3 7 4
the third least significant bit of the RGB values may be 4 4 4
modified. With this, the upper range limit was extended to 5 4 7
127 during the extraction process. This is also done in order 6 7 7
for the program to distinguish which are non- embedded and 7 7 4
embedded pixels. Embedded pixels have values in the range 8 4 7
from 121 to 127 while the non-embedded pixels are
modified in such a way that its equivalent gray values 3.3 Carrier Frame Selection for Video
become at least 128. This modification was done using the Steganography
equations below used in computing the new RGB values of
the pixel components. For video steganography, the audio data is scattered
throughout the video frames. The binary audio data are
Rnew = Rold+ (128 - Gray value) (2) embedded on either every 10th frame or 2nd frame. This
Gnew = Gold+ (128 - Gray value) (3) depends on the size of the audio data that must be
Bnew = Bold+ (128 - Gray value) (4) embedded for every frame as well as the total number of
video frames. By default, the frame interval for
Where Rnew, Gnew,Bnew = new pixel component values embedding is 10. The total number of frames in the video
Rold , Gold , Bold = old pixel component values is used to determine how many frames, in intervals of 10
may be processed for embedding. The audio data is then
The Pixel Intensity Based Selection is done by firswt divided into the number of carrier frames to determine
converting the color images into grayscale. The grayscale how many audio bits must be embedded for every carrier
values are then divided into two groups, values from 0 to 120 frame.
and values from 128 to 255. All pixel colors with an
equivalent gray value of 127 and below are replaced by 0 3.4 Output File Formats
(black) while the remaining upper range is represented by 1
(white). The coordinates of all pixels represented by zero The proposed file formats after the embedding process are
serve as carrier pixel locations. Portable Network Graphics (PNG) for images, Audio Video
Interleaved (AVI) for videos and Waveform Audio (WAV)
3.1.3 Side-most Pixel Selection for the recovered audio data. It is found the in this
implementation, file format MPEG- 4 (MP4) for videos is
In this technique, 80 percent of the total columns in an not compatible for writing the output steganographic video
image or frame are located and assigned as carrier pixels. because it uses standard compression techniques that corrupt
Half of these columns are taken from the leftmost side and or alter the original data.
the other half from the rightmost side. The length of audio
information than can be carried by the cover image or video 3.5 Systems Integration
is limited by the number of carrier pixels of the image or
frame of the cover media. Multiple audio information with The user interface for the system is the android based
variable lengths are used in this research. program used to load or capture the audio information to be

348
Reggie C. Gustilo et al., International Journal of Emerging Trends in Engineering Research, 7(9), September 2019, 346 - 352

transmitted, password generation and extraction and parameters such as MSE and PSNR were computed between
selection of embedding technique to be used. The basic block the embedded file and the received file.
diagram of the system integration is shown in figure 1 below.
The sender and the receiver are linked via a webhost 4. DATA AND DISCUSSIONS
program that overlooks the transmission and reception of the
embedded signal. A server is also used to perform all the Image steganography and video steganography are
complex processing needed in the system such as embedding implemented and tested using the same audio information
of the audio signal into either the image or video cover signals.
media.

4.1 Image Steganography

The data shown to test and implement image steganography

is shown in table 4 below.

Table 4: Data Used for Image Steganography

Height 2835
Width 4252
Image
Size 3.23MB
Row 713726
Channel 2
Audio Frequency 44100
Figure 1: Integrated System Design Size 332KB

Transmission is done in two modes, Bluetooth and In the image steganography, the audio signal is embedded in
email. Steganographic video files are limited to 25MB the image using any of the six embedding techniques that is
for email transmission. chosen by the user. The original image, image with
embedded information and the original audio information is
shown in figure 2.
3.6 Extraction Process

During extraction process, the user must enter the correct

password generated by the system. The system will only
continue if the password entered is similar to the password
generated when the steganographic image or video is
produced. The password was also used to determine details
for extraction, such as the technique used and the number of
bits to be extracted by the system. The bits extracted were
used to reconstruct the audio information that can be
retrieved by the user.

3.7 Evaluation of the system a. Original Image Used for Image Steganography

Two comparative methods were used to measure the

effectiveness of the system. Quantitative analyses were done
with respect to error parameters such as the BER, MSE and
PSNR. The MSE and PSNR were obtained from the
embedded steganographic images and videos compared to
their original unembedded input files. The bit error rates
were obtained by comparing the original and the recovered
audio files.

Qualitative analyses were also done by examining the visual b. Steganographic Image Using Technique 1
changes between the original file and the produced
steganographic file. In support the qualitative method, a
survey was conducted involving 85 participants by asking
them to determine whether or not visual and audible changes
were noticeable. The produced steganographic files were
transmitted using Bluetooth or E-mail. Error checking
349
Reggie C. Gustilo et al., International Journal of Emerging Trends in Engineering Research, 7(9), September 2019, 346 - 352

Table 7: Input Details for Video Steganography

Height 1080
Width 1920
Total Frames 97
Video Frame Rate 24
Duration 4.046
Size 1.55 MB
Row 1764352
c. Original Audio Message Audio Channel 2
Audio Frequency 44100
Figure 2: Original image, embedded image and audio signal used
in Image steganography Size 621KB

Figure 2a shows the original image and the embedded image In video steganography, the six embedding techniques are used to
in figure 2b after using embedding technique 1 using the test the performance of the system. Table 8 shows the password
generated by each technique.
audio signal shown in figure 2c. there is very little or almost
no difference between the original image and the embedded Table 8: Technique and Password Details
image using the naked eye. Technique Password
The password generated during embedding and the 1 1j03z&4H~E
qualitative comparison of the original image and the
2 2j03z&4H~E
embedded image are shown in table 5 below.
3 3j03z&4H~E
Table 5: Qualitative Analysis using Image Steganography 4 4j03z&4H~E
Sample Tech Password MSE PSNR (dB) 5 5j03z&4H~E
1 1 D1j38v#5I| 0.7837 49.2229 6 6j03z&4H~E
2 2 D2j25y$6O+ 2.2569 44.6295
The embedded videos are transmitted wirelessly using
3 3 D3j38v#5I| 1.2384 47.2359
Bluetooth and email. The average performance of the
4 4 D4j25y$6O+ 2.5444 44.1088
transmission is shown in table 9.
5 5 D5j38v#5I| 0.7228 49.5740
6 6 D6j25y$6O+ 1.9691 45.2219 Table 9: Average Transmission Performance for
Video Steganography
From table 5, it can be seen that using image steganography, Technique MSE PSNR (dB)
the original image and the embedded image are visually 1 0.1176 57.4609
similar but qualitatively different as shown by the MSE and 2 0.3351 52.9128
PSNR values that proves the difference in the pixel values of 3 0.1429 56.6133
the two images. 4 0.3588 52.6167
5 0.1168 57.4903
After the extraction and recovery of the audio information 6 0.3344 52.9216
signal, the BER during transmission is measured to check if
the images are altered during the wireless transfer.
Lastly, the original audio messages were compared to the
Afterwards, the original audio data and the recovered audio
recovered audio messages and were tested for accuracy.
data are compared by evaluating the accuracy of the
Table 10 shows the average accuracy of the recovered audio
recovered audio data. Table 6 shows the error checking of
messages.
the audio signal.
Table 10: Audio Error Measurements
Table 6: Audio Error Checking
Technique BER Accuracy
Sample Tech BER Accuracy
1 0 99.5656
1 1 0 99.99998741
2 0 99.5656
2 2 0 99.99998741
3 0 99.5656
3 3 0 99.99998741
4 0 99.5656
4 4 0 99.99998741
5 0 99.5656
5 5 0 99.99998741
6 0 99.5656
6 6 0 99.99998741
5. CONCLUSION
4.2 Video Steganography
A complex android driven system was developed to
The details of the data used for video steganography is
implement image and video Steganography. The user
shown in table 7 below.
interface was driven an android application that was able to

350
Reggie C. Gustilo et al., International Journal of Emerging Trends in Engineering Research, 7(9), September 2019, 346 - 352

load or record the audio signal to be transmitted, choose the wavelet transform, TENCON 2014 - 2014 IEEE Region
type of embedding technique to be used and generate the 10 Conference, 2014.
password for added security during transmission. On the 7. R. A. A. Masilang, M. K. Cabatuan, E. P. Dadios, and L.
receiver side, this android application can extract and recover G. Lim, Computer-aided BSE torso tracking
the embedded audio signal with at least 99.99% accuracy. algorithm using neural networks, contours, and edge
The length of the audio signal is variable and the duration features, TENCON 2014 - 2014 IEEE Region 10
depends on the size of the image cover media or the video Conference, 2014.
cover media. 8. Del Espiritu, J., Rolluqui, G.,Gustilo, R.C., Neural
network based partial fingerprint recognition as
Matlab is used to perform all the complex tasks such as the support for forensics, 8th International Conference on
actual embedding, extraction and recovery of the audio Humanoid, Nanotechnology, Information Technology,
signal. The android program and the Matlab program are Communication and Control, Environment and
linked using a webhost and a server that process and saves Management, HNICEM 2015
all the data of the system. 9. Abinoja, D.D., Roque, M.A., Atienza, R., Materum, L.,
Landmark-based audio fingerprinting algorithm for
The embedded images and videos are transmitted wirelessly a transmitter-less alert recognition device for the
via Bluetooth and email. BER and PSNR are used to hearing-impaired, 8th International Conference on
quantify the performance of the system during the Humanoid, Nanotechnology, Information Technology,
transmission stage. Communication and Control, Environment and
Management, HNICEM 2015, .
Experiment results show that the presented steganography 10. A. Africa, A Rough Set Based Solar Powered Flood
methods are proven to be efficient and effective way of Water Purification System with a Fuzzy Logic
concealing audio information inside images or videos. Model. ARPN Journal of Engineering and Applied
Sciences. Vol. 12, No. 3, pp.638-647, 2017
REFERENCES 11. A. Africa and J. Velasco, Development of a Urine
Strip Analyzer using Artificial Neural Network using
1. Gustilo, R.C., Dadios, E.P., Machine vision support an Android Phone. ARPN Journal of Engineering and
system for monitoring water quality in a small scale Applied Sciences. Vol. 12, No. 6, pp. 1706-1712, 2017
tiger prawn aquaculture, Journal of Advanced 12. S. Brucal, A. Africa, and E. Dadios, Female Voice
Computational Intelligence and Intelligent Informatics, Recognition using Artificial Neural Networks and
20(1), pp. 111-116, 2016 MATLAB Voicebox Toolbox. Journal of
https://doi.org/10.20965/jaciii.2016.p0111 Telecommunication, Electronic and Computer
2. Gustilo, R.C., Dadios, E.P., Behavioural response Engineering. Vol. 10, Nos. 1-4, pp. 133-138, 2018.
analysis using vision engineering (BRAVENet), 13. Villanueva, L., Gustilo, R.C., Artificial neural network
Journal of Advanced Computational Intelligence and based antenna sensitivity assignments for chaotic
Intelligent Informatics, 21(2), pp. 211-220, 2017 Internet Service Provider network architecture,
https://doi.org/10.20965/jaciii.2017.p0211 International Journal of Engineering and Technology
3. E. Mohammadi, E. P. Dadios, L. A. G. Lim, M. K. (UAE), 7(2), pp. 14-17, 2018
Cabatuan, R. N. G. Naguib, J. M. C. Avila, and A. 14. Dulay, A., Sze, R., Tan, A., Yap, R., Materum, L.,
Oikonomou, Real-Time Evaluation of Breast Self- Development of a wideband PLC channel emulator
Examination Using Computer Vision, International with random noise scenarios, Journal of
Journal of Biomedical Imaging, vol. 2014, pp. 1–12, Telecommunication, Electronic and Computer
2014. Engineering, 2018
4. R. A. A. Masilang, M. K. Cabatuan, and E. P. Dadios, 15. Hanpinitsak, P., Saito, K., Takada, J.-I., Kim, M.,
Hand initialization and tracking using a modified Materum, L., Multipath clustering and cluster
KLT tracker for a computer vision-based breast self- tracking for geometry-based stochastic channel
examination system, 2014 International Conference on modeling, IEEE Transactions on Antennas and
Humanoid, Nanotechnology, Information Technology, Propagation, 2017
Communication and Control, Environment and https://doi.org/10.1109/TAP.2017.2754417
Management (HNICEM), 2014. 16. Abinoja, D.D.N., Materum, L.Y., BIC-based
5. M. K. Cabatuan, E. P. Dadios, R. N. Naguib, and A. optimization of the identification of multipath
Oikonomu Computer vision-based breast self- propagation clusters in MIMO wireless systems,
examination palpation pressure level classification ISAP 2016 - International Symposium on Antennas and
using artificial neural networks and wavelet Propagation, 2016
transforms, TENCON 2012 IEEE Region 10 17. Dulay, A.E., Yap, R., Materum, L., Hardware
Conference, 2012. Modelling of a PLC Multipath Channel Transfer
https://doi.org/10.1109/TENCON.2012.6412282 Function, Journal of Telecommunication, Electronic
6. J. A. C. Jose, M. K. Cabatuan, E. P. Dadios, and L. A. and Computer Engineering, 2017
G. Lim, Stroke position classification in breast self- 18. Materum, L., Stochastic tapped delay line based one-
examination using parallel neural network and sided beamformed channel impulse response models
351
Reggie C. Gustilo et al., International Journal of Emerging Trends in Engineering Research, 7(9), September 2019, 346 - 352

of LoS and reflected waves at 62.5 GHz in a https://doi.org/10.30534/ijatcse/2019/14832019

conference room environment, Journal of 20. Kaushik H. Raviya, Dr. Dwivedi Ved Vyas and Dr.
Telecommunication, Electronic and Computer Ashish M. Kothari, SVD Based Performance
Engineering, 2017 Improvement in Hiding a Message Behind an Image,
19. Monica Thomas and Dr. Varghese S Chooralil, Security International Journal of Advanced Trends in Computer
and Privacy via Optimised Blockchain, International Science and Engineering, Volume 8, No. 2, 2019
Journal of Advanced Trends in Computer Science and https://doi.org/10.30534/ijatcse/2019/12822019
Engineering, Volume 8, No. 3, 2019

352