CHAPTER 1

INTRODUCTION
This chapter explains in brief about the steganography includes classification of steganography and
also the objective, methodology of the proposed work.

Steganography is a method of protecting essential information or messages from unauthorized users

by hiding them inside other carrier data. Because the Human Visual System (HVS) will not be
capable to distinguish slight variation that happens in the cover file, the mixed data, also known as
stego objects, will be perceived as one part of data by the HVS. Text, audio, image, and video can be
used in the message and cover data [1]. As illustrated in Figure 1.1, the fundamental steganographic
system consists of several components, including cover media, message, and the algorithm, which is
the most important component. [2]. Any efficient steganography system should take into account the
three essential criteria: embedding capacity, embedding efficiency or imperceptibility, and robustness
against attacks [3].

Figure 1. 1: Basic steganography system

To start with, the embedding payload is the quantity of concealed data that will be encoded within
the cover file. The approach has a high embedding payload when the secret message has a large
capacity [4]. The visual quality of the stego, as well as its security and robustness to attackers, are all
considerations in embedding efficiency. Second, both a truncated modification rate and worthy cover
data acceptable quality result in a high imperceptability. The high embedding efficiency of the
steganography technique will limit invader thought of judging secreted data and will be hard to
perceive using steganography tools [1]. Any alteration to the cover files following the embedding
process, on other hand, will draw attackers' attention. The steganographic scheme's security has a
direct impact on embedding efficiency. Increasing the secret message's capacity lowers the quality of
stego videos, which reduces embedding efficiency. Both considerations should be taken into account
[3].

1.1 GENERAL CLASSIFICATION OF STEGANOGRAPHY

 Hidden Data : The data that needs to be embedded or that needs to be hidden from everyone
except the intended receiver.
 Cover Media : The data will be integrated in this medium. An image, an audio file, or a video
file can be used as a cover/carrier media file.
 Stego Media: Cover media that has the secret information [5].

Figure 1. 2: Classification of steganography

The study of undetectable communication is known as steganography. This information is transferred
by concealing data within media. Various steganographic approaches can be classified into four
primary components, as indicated in Figure 1.2. They are as follows [2]:
 Image Steganography : To conceal the hidden message, we employ an image file as cover
medium. A digital image is made up of low and high frequency information. A low frequency
zone has a strong relationship with its neighbour pixels , while a high frequency region has a
strong separation from its neighbour pixels. The hidden message is placed into the image pixels
using several transform techniques to disguise the data in an image, taking advantage of human
eye sensitivity. Recently, a few machine learning techniques have been utilized to improve the
robustness, embedding capacity, and other aspects of the system [1].
 Audio Steganography : It's a method of sending a concealed message using an audio file. Since
human auditory system (HAS) is highly sensitive than human visual system (HVS), embedding
the data in an audio file is much highly harder than encoding a secret data in an image file.
Because the message is encoded in audio signals, approaches such as LSB coding, parity coding,
and echo data concealing are utilised to embed the procedure in an audio file [2].
 Text Steganography : It deals with modifying the format of a file's existing text, changing the
words within the text, and producing random character sequences. We're simply using the text
file as a cover media to conceal the hidden information in this instance. Because an attacker may
easily find the pattern, it is more vulnerable to attack [5].
 Video Steganography : To conceal the hidden message, a video file is employed as a cover
media. Because a video file combines text, image, and audio, it is less vulnerable to steganalysis.
It is a sequence of frames that are played again at a consistent speed and measured in frames per
second. We must first extract the frames from a video file in order to put a secret information in
it. Frame conversion is done first in order to incorporate the information in the cover video file.
It's a method of transforming a video into subsequent images or frames, then using each or one
of those frames as carrier frame to hide concealed message. To create stego video, all frames are
blended together after embedding procedure. Spatial and transform domain techniques are used
in video steganography [1, 2, 6].

1.1.1 SPATIAL DOMAIN METHODS

Least significant bit (LSB) replacement, Pixel Value Differencing (PVD), Region of Interest (ROI),
histogram manipulation, matrix encoding, and mapping rule [4] are just a few of the steganographic
methods using spatial domain. One of the most basic methods is to use an LSB substitution approach
to embbed information in a video file [7]. Data is sequentially stored in all pixels of a frame in this
technique. If a hacker knows the frame number [8], he can extract the message with ease. Two
sequential pixels are chosen in PVD, and these pixels are implanted with the information. The
difference between these two successive pixels is used to determine payload and used as base for
finding if two pixels belong to a smooth area or an edge In order to improve the capability of PVD
while also improving the technique's quality, a process that combining PVD and LSB is presented
[9]. The LSB approach is a good means to hide data, but it has a flaw in that it is easily detected by
an intruder, hence PVD is used to fix this problem.

1.1.2 TRANSFORM DOMAIN METHODS

The secret information is encoded using the low, middle, or high frequencies of the transformed
coefficients in transform domain steganographic methods, and each video frame is individually
transformed into frequency domain using discrete cosine transform (DCT), discrete wavelet
transform (DWT), and discrete fourier transform (DFT). In this field, many methods for concealing
information in a frame have been presented. One of these is to use wavelet transformations, which
divide a video frame into four sub-bands and separate the odd and even coefficients. After that, secret
data is encoded in the RGB coefficients. The length of concealed information is hidden into
coefficients for reliable embedded data extraction [8]. These methods have a large embedding
capacity, however after embedding the video, they have a lower average video clarity [3, 9].

1.2 NEED FOR RESEARCH

Steganography is a process of secret communication in which presence of hidden information is
concealed. Any digital entity, such as a text file, image, audio, or video, can be used as the cover
medium for steganography With TCP/IP packets (Transmission Control Protocol/Internet Protocol).
Making use of video because digital is widely used as a cover media for steganography, it is a well-
known practice on the internet, videos are commonly used. The basic goal of steganography is to
conceal information, undetectability, higher embedding capacity, and robustness are all advantages.
It is the better option that a steganography algorithm should be able to withstand both intentional and
unintended eavesdropping during transit, the enemy attempted image alteration. There is no single
steganography algorithm that can handle and solve all of the above-mentioned needs (Tayana Morkel
et al 2005). Furthermore, the steganography objectives are interconnected. It cannot be accomplished
by using a single steganography technique and a certain image format. All of the aforementioned
objectives are met in the proposed study by using a video in the spatial domain. Geometrical
distortion, such as rotation/flipping, can be determined and remedied in the case of robustness if any
occurred accidently in transit or produced incidentally by the intruder. The objectives are reasonable
and sufficient for stamping the method created in the proposed study. In various ways, the
performance achieved using the algorithm created in the proposed study was superior to that of many
other algorithms.

1.3 MOTIVATION
India is the world's second-largest victim of cyber-attacks, with a 40 percent increase expected by
2020.
As a result of the increased use of the internet, security is receiving a lot of attention. As the number
of people using the internet grows, so does the amount of data that is shared each day. Hackers may
target the data that is exchanged on a daily basis. In order to deal with this issue Steganography is
one of the most effective solutions for dealing with this problem.
This motivated for development of video steganography techniques for data hiding with enhanced
security.

1.4 OBJECTIVES
 The project mainly focuses on Enhanced security by hiding the data using video steganography
techniques.
 To increase the capacity, imperceptibility and robustness of the cover media.
 To calculate and also to compare the parameters that is PSNR, MSE and SSIM using
• Neural Networks using hash function and
• Modulo operator based techniques.

1.5 PROBLEM STATEMENT

 Eavesdropping and unauthorised access to information communicated via the internet have
harmed data confidentiality and integrity. As a result, information security becomes necessary.
 Data can be concealed in carriers like images, audio files, text files, and video files using
steganography.
 The proposed video steganography techniques aims at a secure authentication system based on
hiding the data like Images and text extracted from video.
1.6 METHODOLOGY
In the proposed methodology, two different video steganography techniques are used and both are
compared based on the results obtained.

 Video steganography using modulo operator.

Secret text and images are placed as secret information within a cover video in this method. The
secret data is pre-processed before being placed using modulo operator during the embedding
procedure.
It includes two stages,
A. Embedding
1) Pre-processing
2) Embedding using modulo operator:
B. Extraction
1) Extraction using modulo operator:
2) Post-processing
 After extraction, parameters are calculated that is PSNR, MSE and SSIM by the result obtained.

 Video steganography by neural networks using hash function.

Hybrid neural networks and the hash technique are used to hide secret information in a video file
in this method.
The approach has two significant steps:
1) Encryption or Embedding process
2) Decryption or Extraction process
A. Embedding process
Embedding, also known as the Encryption technique, is a process in which the cover video is
combined with the secret input to generate a stego-object.
B. Extraction process
Extraction, often known as decryption, is the procedure of obtaining the hidden image from a stego
video.
 After extraction, calculate the parameters that is PSNR, MSE and SSIM by the result obtained.
1.7 ORGANIZATION OF THE REPORT
 Chapter 1: This chapter comprises of brief introduction about steganography and also explains
about motivation and problem statement. This also comprises the brief overview of the project.

 Chapter 2: This chapter converse about the history and the related work carried out or proposed
earlier by researchers and engineers.

 Chapter 3: This chapter describes the proposed work, its working and a detailed information
about the algorithm used in the work.

 Chapter 4: This chapter discuss about the system and software requirements.

 Chapter 5: This chapter discuss about the results obtained for each step and also performance
analysis in the project.

 Chapter 6: The conclusion, applications and the possible future work of the proposed work is
discussed in this chapter.

This chapter deals with the brief explanation of video steganography techniques in data hiding.
This chapter also gives an introduction about security while hiding the data and improves the
performance level on video code streams without compromising the video quality using video
steganography techniques which are going to be implemented further in this research work.
CHAPTER 2
LITERATURE SURVEY

This chapter describes about the background history of steganography, some of the related research
papers and also the conclusion drawn from the review papers.

2.1 HISTORY OF STEGANOGRPHY

Steganography is being used over a long time to exchange, communicate in a secure manner. Some
examples are:
 This technique was first utilized in ancient Greece around the fifth century. In Greece, people
used to hide information on the heads of their slaves. Slave was picked first, and his head was
shaved before a message was put on it. They waited till the hair turned dark and the message was
hidden before proceeding. That slave was transferred to another location, where his skull was
shaved once more in order to obtain a secret message [10].

 Spartans adopted the Steganography technique at the same time as the Greeks. This approach
uses the employed and Demerstus forms. Against Xerxes, the Spartans employed this strategy. A
hidden information was carved on a wood wax tablet and shielded with a new plane layer of wax
because the wax appeared to be blank [11].

 Sir Francis Bacaon uses the steganography technology in a face variant encoding in the 1600s.

 During World War II, the steganography technology was utilised to conceal the existence of data.
Invisible ink was used to write information on paper, which made the paper appear blank to a
normal human under normal light. Finally, liquids such as water, fruit juices, and vinegar were
used to read out the information. First, the moist paper in liquid was heated due to which they
turned black and information written using invisible ink made visible to eyes [12].

 During World War II, Germans created a new steganography technique known as microdots, in
which a secret message was disguised in an innocent message in such a way that the letter at each
certain place in the message indicated the letter of the secret message [13].

 Morse code was hidden in a picture in 1945. The coded message is embedded on the grass near
the river's edge. A line was signified by long grass, while a point was denoted by short grass.
“Compliments of CPSA MA to our chief Col Harold R. Shaw on his visit to San Antonio May
11th 1945” [14] read the decoded message.
2.1.1 THE DIGITAL ERA OF STEGANOGRAPHY
Because of developments in computer power, the internet, and the emergence of digital signal
processing, information theory, and coding theory, steganography has gone "digital". Steganography
has generated an atmosphere of corporate vigilance in this digital age, which has fostered a variety
of innovative applications, ensuring its continued progress. This digital revolution is thought to
benefit cybercrime. As a result of a revelation in USA TODAY, there was an immediate outpouring
of worry about criminals' potential use of steganography. It's tough to regulate cyber-planning, or the
"digital scourge," as Lieutenant Colonel Timothy L. Thomas put it. Three million photographs from
popular websites were examined by Provos and Honeyman for any steganographic evidence [15].
They haven't discovered any hidden messages. Despite the fact that they blamed a variety of factors
for the failure, it's worth noting that steganography isn't limited to still images. Hidden messages can
also be embedded in video and audio files.

2.1.2 STEGANALYSIS
Steganalysis is a type of steganography attack. In essence, it is a technique for noticing the presence
of a secret message in text/image [16]. It compares the messages and looks for the concealed message.
When a message is disguised in an image, the intensity may be slightly reduced and the colour may
be slightly faded, as we all know. As a result, this aids in the detection of any hidden messages.
Steganalysis, as we all know, aids in the attack on steganography. The following are some of the most
common forms of attacks:

i. Known carrier hack

ii. Only steganography attack

iii. Only known message hack

iv. Only known steganography hack

In the spycraft family, steganography and cryptography are cousins. Cryptography scrambles a
message by turning secret data into incomprehensible form using cryptographic methods [16].
Steganography, on the other hand, hides the message and makes it unreadable. Steganography, to put
it another way, is science of concealing secret information. Cryptography's aim is to keep information
unreadable by an unauthorized person, whereas steganography's objective is to keep data hidden from
an unintended spectator.
2.1.3 CRYPTOGRAPHY
Cryptography is made up of two Greek words: crypto (meaning "secret") and graphy (meaning
"writing"). So, cryptography is a process of converting a message/information from one form to
another secret form that is different from the original with the use of a secret key [17]. We, on the
other hand, using a term Cryptography to describe science and art of changing signals to make them
more secure and resistant to assault. Cryptography is the process of changing the value of a secret
message, while decryption is the process of recovering the original message from the cipher.

Figure 2. 1: Basic Cryptography process

2.2 COMPARISON OF DIFFERENT SECURITY SYSTEMS

The concept of "What You See Is What You Get, WYSIWYG," which is at times used when printing
photographs or other materials, no longer exact and not able to mislead a steganographer because it
is not correct every time. Videos can transmit more than 1000 words since they can be more than
what the Human Visual System, HVS, can see. People have spent decades trying to come up with
new ways to communicate secretly. Steganography, watermarking, and cryptography are three
techniques that are related. The first two are particularly difficult to distinguish, especially for those
with backgrounds in other disciplines [18].
Drawing a distinction between these methods is both arbitrary and perplexing [19]. As a result, before
providing a detailed analysis of these strategies, it is necessary to address them briefly.

Table 2. 1: Comparison of Steganography, Cryptography and Water marking

Figure 2.2 and Table 2.1 may help to clear up any ambiguity. The study given here focuses on
steganography in digital photographs and excludes other kinds of steganography like linguistic or
audio steganography.

Figure 2. 2: The different embodiment disciplines of Information Hiding.

Figure 2.3 indicates that media TV networks frequently have their logos watermarked for their
broadcasting. Table 2.1 outlines variations and similarities between steganography, watermarking,
and cryptography. Throughout this dissertation, the phrase "cover image" refers to the image that will
convey the embedded bits. The term "stego-video" refers to a video with embedded data (payload),
whereas "steganalysis" or "attacks" refers to various image processing and statistical analysis
methodologies aimed at breaking steganography algorithms.
 Figure 2. 3: Media TV channels usually have their logos watermark

2.3 VIDEO STEGANOGRAPHY LITERATURE SURVEY

Kiranjit kaur and Baldip kaur [20] used a decomposition technique with a hybridization of DWT
and LSB in the proposed work to locate the optimal bits from a cover video to include the secret
information without sacrificing quality. This research focuses on enhancing the rate of security in
practical applications by combining decomposition techniques with machine learning methods. ANN
has been employed in this case for the enhancement in hiding process.

Anamika Saini, Kamaldeep Joshi and Kirti Sharma [21] described a review and analysis of video
steganography technique is applied to an AVI (Audio Video Interleaved) file using the LSB (Least
Significant Bit) technique, and frames are compared, analyzed using metrics like PSNR and MSE.
The major limitation of this approach is that the LSB technique is easy to decode, making it vulnerable
to attacks.

Ramadhan J. Mstafa and Khaled M. Elleithy [22] describes a comprehensive research study and
exploration of different cutting-edge video steganography techniques, as well as their implementation
evaluation. This evaluation offered a flow inquiry into bearings and actions to improve existing video
steganography systems.

Sushmita MC, Suresh HN & Manikandan J [23] using an ideas of discrete wavelet transform and
steganography, a novel attempt to hide one or two secret videos in the single cover video was
developed. The PSNR values of the stego-video and recovered concealed videos are acceptable for
visualisation with human naked eyes, with no reduction in video quality.
Hemant Gupta and Dr Setu Chaturvedi [24] provides a sophisticated data hiding approach based
on LSB substitution and analyses the correlation between the original and embedded image for 1bit,
2bit, and 3bit LSB Substitution. If there is no modification in an original image, hence correlation
coefficient has the value r=1. When a number of LSB Substitutes increases, correlation factor
decreases. The PSNR (peak signal-to-noise ratio) between an original image and an encrypted
(received) image was also calculated in this paper. As a result, the number of LSB Substitutes
increases, hence PSNR decreases, and also data security increases at the same time.

Mritha Ramalingam and Nor Ashidi Mat Isa [25] in the IWT domain, a method was created to
conceal data in a grey scale frame of a video. The data are concealed in all frequency bands using this
method. For data concealment and extraction in Red Green Blue (RGB) components of video files,
this work provides a video steganography technique based on Haar Integer Wavelet Transforms
(IWT) and Least Significant Bits (LSB) replacement. The cover-video is separated into RGB frames
in this method, and text is inserted in binary form in LSBs of IWT coefficients. Using a reverse
technique of data concealing, the embedded text is recovered from stego-video. Audio Video
Interleave (AVI) files were used to implement the proposed system.

Sachin Jangid, Somesh Sharma [26] proposes a video Steganography Algorithm based on MLC
(Multi Level Clustering). For clustering the Cover Frame, the Multi-Clustering Algorithm employs
K-Mean clustering. LBP will insert a secret message into selected clusters (Local Binary Pattern).
The performance of Video Steganography is evaluated in terms of PSNR (Peak Signal to Noise Ratio)
and MSE (Mean Square Error) (Mean Square Error).

S. D. Hu and U. K. Tak [27] presents a unique Video Steganography technique for concealing an
uncompressed secret video stream within a host video stream of nearly equal size. Non-uniform
rectangular partitioning will be used to divide each frame of secret video, and partitioned codes
acquired can be an encrypted version of the original frame. These codes will be encoded in the host
video's Least 4 Significant Bits of each frame. Experiments revealed that technique can covert a video
of the same size in host video without causing noticeable distortion.

K. Dasgupta, J. K. Mondal and P. Dutta [28] introduces a revolutionary video steganography

system for information concealment that is both efficient and effective. For video steganography, a
3-3-2 LSB-based system was utilised as the foundation. As a result, the fundamental technique is
improved by applying a Genetic Algorithm (GA), which strives for optimal hidden data
imperceptibility. To determine the frame's innocence in relation to the original, an anti-steganalysis
test is utilized. The optimizer optimises the values over a 3-3-2 LSB approach for basic video
steganography. The optimizer utilises a cost function with two factors, but extra factors can be added
for a more in-depth analysis. On the basis of perceptibility and fidelity, proposed technique was
compared to the base techniques in terms of performance. The strategies are used in uncompressed
domain.

S. Bhattacharyya, A. Khan, A. Nandi, A. Dasmalakar, S. Roy and G. Sanyal [29] presents a

method for image data steganography based on the pixel mapping method (PMM) and bit-plane
complexity segmentation (BPCS). The proposed method works by employing a mathematical
function to pick the embedding bit planes, and then using the pixel mapping method (PMM) in 8x8
blocks of each selected plane. The combined approach of PMM and BPCS results in a reliable image-
based steganography method that is unaffected by the nature of data to be buried and produces a stego
image with minimal degradation.

2.3 CONCLUSIONS DRAWN FROM LITERATURE SURVEY

 There are numerous embedding algorithms in the spatial and transform domains, as evidenced by
existing techniques. When compared to transform approaches, spatial domain techniques had
greater PSNR values but were less resistant to attacks.
 Many different approaches to embedding were used in this spatial domain technique, such as LSB
or histogram, however these strategies were unable to preserve the secrecy over time in stego-
media.
 Transform domain approaches are more secure since it is difficult to determine the embedded
pixels in the cover material.
 These approaches have different bands, and the embedding capacity is high. However, the PSNR
of these approaches is modest.
 As a result, we require a different spatial domain approach that is difficult for the attacker to
exploit while maintaining a high PSNR value.

By considering all the conclusions the proposed Neural network with hash function algorithm can
help in increasing the robustness to attacks as the embedded pixels will be scattered among the frame
which makes it difficult for the attacker to decode. Here another Modulo operator technique is also
proposed and compared with neural network using hash function technique.
CHAPTER 3
METHODOLOGY
This chapter describes about the methodology of two video steganography techniques proposed and
also how it was implemented in the project. In the proposed methodology, two different video
steganography techniques are used and both are compared based on the results obtained.

3.1 VIDEO STEGANOGRAPHY

Video steganography is a technique for hiding data in a video file as it is being created. Because of
its size and memory requirements, video steganography is more appropriate than multimedia files. A
digital video including a set of frames cooperating back at determined frame rates based on the video
standards is known as video steganography. The information in certain individual frames is hidden
via video steganography. It's difficult to see which frame the data or information are hidden in after
they've been covered. In the video approach, steganography is divided into two distinct procedures.
The first is to embed the data in unprocessed video. The next step is to attempt to incorporate data
directly into a corrupted video stream.

Figure 3. 1: Block diagram of video steganography

Various plans are used in video steganography. The optimum strategy is to keep the covert
information hidden while lowering the aspects of the cover video such that it is impossible to detect
with the naked eye. The dispatcher recognizes the attached video as the "stego" video and sends it to
the customer's region. Essential files are protected using a combination of video steganography
models.
The primary goal of steganography is to hide existence of data while maintaining quality of cover
object. Figure 3.1 depicts the basic processes involved in steganography (image/video) method. The
cover image or video is pre-processed and split into frames at sender's end, and secret information is
encoded in image or video frames during embedding process, resulting in a stego-object (image or
video) to transmit across a channel. The stego-object is utilized on the receiver's side to extract the
secret information using the extraction procedure, and same pre-processing is used to re-create cover
image or video. The illustration depicts block diagram of reversible approach, which means that after
extracting the secret data, the entire cover image or video is recreated at the receiver's end.
As stated in chapter one, steganography techniques are divided into two categories: spatial domain
methods and transform domain methods. While a transform domain technique is safer, a spatial
domain approach allows for greater payload embedding. The simplest data hiding technique in the
spatial domain approach is the least significant bit (LSB) replacement method, in which the pixel's
least significant bit is replaced the value of secret data value. There are numerous tools in the
transform domain, including the Fourier Transform (FT), Discrete Cosine Transform (DCT), and
Discrete Wavelet Transform (DWT). Apart from these two types of steganography, there is also
adaptive steganography (AS). It can be used in both spatial and frequency domains. It's a unique
opportunity.
In most cases, data embedding is done directly on the data, or after the data has been decomposed or
transformed. Although there are many different data embedding strategies, they are all subject to
attacks, which implies that data security may be affected. We can apply machine learning approaches
in data embedding to solve this problem.
The fundamental idea behind this project is to utilize a neural network with a hash function method
to embed the message in stego frames. A Modulo operator approach is utilized to compare the
findings in addition to the neural network approach.

3.2 VIDEO STEGANOGRAPHY USING MODULO OPERATOR

The usage of the modulo operator has the drawback of causing overlapping during the extraction
process. When the remainder of a different dividend is divided by the same divisor, the same result
can be obtained. By dealing with this issue, the suggested technique aims to reduce data loss at the
receiver end. Within a cover video, secret data is embedded as target data. The data is pre-processed
before being inserted using the modulo operator during the embedding process. Figure 3.2 shows the
modulo algorithm flow chart.
It consists of two stages:
 Embedding process
 Extraction process

Embedding process Extraction process

Figure 3. 2: Flow chart of proposed modulo based steganography

A. Embedding: Embedding, also known as the Encryption method, is a procedure in which the
cover video is combined with the secret input to produce a stego-object.

1. Preprocessing: For embedding the secret data inside a cover video, it has to be preprocessed into
cipher form to embed it with the cover frame.
 For secret text preprocessing, each character of the target string is converted to their
corresponding ASCII values, which are then converted to binary representation.
 For preprocessing of secret image, resize the image that will be lesser than cover frame size and
reshape it into a single row matrix form.
2. Embedding using modulo operator: In embedding process, the cipher text obtained by
preprocessing that is binary digit values will be embedded into cover video using modulo
operation. After embedding stego video file is obtained that will be send to the receiver.
 In modulo operation, consider three variables p, h and r
Where, p = 2^𝑏 (bit value where secret data bits has to hide)
h= 2^(𝑏−1) (previous bit value)
r = remainder
• The pixel values are divisible by p, then if the remainder (r) >= h then the pixel value is subtracted
by h.
• For eg, consider pixel value of cover frame I(k) = 184. If the secret bit is hiding in high LSB bit
that is 4th bit, then p=16, h=8.
184 = 10111000 (here 4th bit is high)
• For embedding secret data bit in 4th bit it should be ‘0’. Hence modulo operation is applied,
if (I(k)/p >= h), then
I(k) = I(k)-h
i.e, 184/16 = 8, then I(k) = 184-8 = 176 (pixel value becomes 176)
176 = 10110000 (4th bit becomes 0)

Algorithm for embedding

Step 1: Input the cover-video which is selected for the encryption process. The video is then
converted into frames. The number of frames obtained depends on the video which is been selected
for this process. For the implementation, the secret key that is frame number in which secret data has
to be hidden is being shared personally, so the sender and the receiver have the frame number with
them.
Step 2: The secret data that we're encrypting in the video is first transformed to cypher form, and
then the video frame in which the message is to be concealed is transformed to RGB components.
Step 3: Extract the length of the secret message, then extract the pixel values of the cover frame that
are identical to the length of the secret message, and save the remaining pixel values in a matrix.
Step 4: Read the secret data bits to be encoded only in the red frame of the image because changes
in this component will not effect the final output stego-video.
Step 5: Use the modulo operator technique on the red frame for encoding secret data.
Step 6: Finally all the pixel values after the embedding process is converted back to the frame by
reshaping it into the matrix form.
Step 6: Finally the stego frame is placed back in the video and stego video is formed.
Step 7: End.

B. Extraction: Extraction, often known as decryption, is the process of obtaining the hidden image
from a stego video.

1. Extraction using modulo operator: At the receiver side, from the stego video file obtained by
the sender secret data will be extracted using modulo operation. The secret message is in cipher form.
This is now sent to the post-processing approach in order to obtain the original target data.
2. Post-processing:
 The generated binary digits are translated to their decimal equivalent for post-processing of
secret text. To acquire the original text, the characters of these ASCII values are concatenated
according to the string length.
 For post-processing of secret image that is in matrix form will be reshaped to image, then the
original image will be obtained.
 After extraction, parameters are calculated that is PSNR, MSE and SSIM by the result obtained.

Algorithm for Extraction

Step 1: Input the stego-video which is received for the decryption process. Then this video is
converted into frames. The number of frames obtained depends on the video which is been selected
for this process.
Step 2: The video frame is selected in which the message is hidden, these stego-video frame is
decomposed into RGB components.
Step 3: Obtain the pixel values of the frame and extract the data or message using modulo operation.
Step 4: To acquire original target data, post-process the extracted data.
Step 5: Finish.
3.3 VIDEO STEGANOGRAPHY BY NEURAL NETWORKS USING HASH
FUNCTION.
Using the hash algorithm, hybrid neural networks are utilized to hide secret information in a video
file. Figure 3.3 depicts the proposed method's flow chart. In our suggested technique, the video is
broken down into number of frames, and random frame is selected to conceal secret information
using secret key, and bit position to encode the secret information is retrieved using a hash function
that is learned using Neural Networks, which improves security.

Embedding process Extraction process

Figure 3. 3: Flow chart of proposed neural network using hash function based
steganography
Hybrid Neural Networks are a combination of Convolutional and Recurrent Neural Networks that
are used to classify pixels in less sensitive areas from pixels in more sensitive parts in order to embed
hidden data.
The two main stages of the procedure: Encryption or Embedding process and Decryption or
Extraction process.

A. Embedding process: The Encryption technique, also known as embedding, is a method of

creating a stego-object by combining the cover video with secret input.
Algorithm for Embedding:
Step 1: First, the AVI-formatted cover video is uploaded. The AVI file is used, because it includes a
sequence of high-resolution images known as frames.
Step 2: Hide the message in any form of media, such as photos, audio, or video.
Step 3: Split the video into frames and use the secret key to encrypt the message.
Step 4: On the chosen frame, perform the following steps: Choose the finest part of the cover video
to embed the hidden information using the hash function and neural networks algorithm. The output
layer provides a trained neural object that is used for categorization.
Step 5: Ultimately, the finished stego-video containing the hidden message is obtained.

B. Extraction process: Extraction, often known as decryption, is the process of obtaining the hidden
image from the stego video.

Algorithm for Extraction:

Step 1: Read the stego video and reverse the approach to retrieve the secret data from the stego video
for the extraction process.
Step 2: Finally, the hidden image that results is received.

3.4 HASH BASED LEAST SIGNIFICANT BIT TECHNIQUE

With the Hash-based Least Significant Bit (HLSB) method for Video Steganography, the secret data
is hide in the LSB of cover frames. This technique is used to encode eight bits of secret data into the
RGB pixel values of video frames in a 3, 3, 2 manner. The insertion point in LSB bits is chosen using
a hash function. For instance, the RGB pixel value for the cover frame is:
R: 10110111 G: 10010100 B: 11001001 and a message byte of 10001001 to be placed in the LSB.
The lowest bit in a series of binary numbers is called the LSB, and in this example it will be 1 for R,
0 for G, and 1 for B. In each pixel value, the proposed approach is applied in the four lowest LSBs.
R: 0111 G: 0100 B: 1001 are the LSBs for the RGB values above.
In the respective RGB LSBs positions, the message is contained in groups of 3, 3 and 2. The hash
function in equation,
k=p % n ………………………………. (1)
is used to calculate the positions. For the current case, n number of LSB bits is equal to 4. The
insertion of secret bits in RGB pixels is as shown below in figure 3.4.
 Figure 3. 4: Embedding secret bits in Hash-LSB method

Let k = 1, 2, 3 for R using hash function be the position of insertion k returned for certain iteration. k
= 4, 1, 2 in the case of G, and 3, 4 in the case of B.
Based on the above insertion points, the bits from message are inserted in four LSB locations, yielding
the RGB pixel value shown below. R: 10111001 G: 10011000 B: 11001001.
As a result, the message's eight bits are encoded in three bytes, with just five of the twenty-four bits
actually changed. Furthermore, these five bits are spread randomly, increasing the scheme's
robustness.
The valid user decodes the message by going backwards in time. The hash function calculates the k
values to determine the insertion place because it is known to the intended user. R: 10111001 G:
10011000 B: 11001001 (using the same embedded RGB value as above).
The following k values will be returned by the hash function for this iteration.
k = 1, 2, 3 for R. k = 4, 1, 2 for G. k = 3, 4 for B. For these k values, which represent four LSB places,
can be used to extract the secret message data as follows: 10001001 which is the same as the secret
message mentioned earlier.

3.5 NEURAL NETWORK (NN)

Machine learning algorithms aid in decision-making, and neural networks have shown to be effective
in categorising data in a variety of fields. The neural network is the most widely used approach among
them. Neural networks are networks made up of simple components that functions together in a
parallel manner.
The NN algorithm will be used to classify the images. A neural network is composed of "neuron-
like" components called nodes, as depicted in figure 3.5. This neural evaluation approach is used to
solve regression problems and is used in the disciplines of optimization, classification, and control
theory.
In classification issues where target detection and recognition are required, NN are particularly
effective. Because of its active nature, NN is favored above other approaches.
This active character is achieved by adjusting the weights based on the final result as well as the
applied input data. The weights are adjusted iteratively until the desired output is achieved.
By altering the weights between nodes, a neural network can be trained to execute a certain purpose.
To generate appropriate output, a variety of activation functions are used.

Figure 3. 5: Neural network

3.5.1 CONVOLUTIONAL NEURAL NETWORK
Convolutional neural network (Conv Net or CNN) is a common deep learning approach, a type of
computer training in which a model learns to execute classification functions directly from images,
video, text, or voice. CNNs are particularly useful for identifying patterns in products, faces, and
images in photographs. It learns from images directly, employing models to rate them and obviating
the requirement to extract particular attributes. Neural networks aren't new, of course. Academic
papers from the 1940s can be found with a quick search. The convolutional flavour, on the other
hand, is more recent and has risen in prominence in recent years as a result of a revived focus on deep
learning. In the past, Conv Nets were employed in the field of computer vision to process picture
data. ConvNets are notable for how well they perform when it comes to detecting picture features
such as people's faces or a self-driving car's constantly updated external view of the environment.
ConvNets are also useful for processing natural language (unstructured data) and optical character
recognition (OCR). The structure of CNN with the layers is as decipted in figure 3.6.

Figure 3. 6: CNN layers

These levels are in charge of actions that include the exchange of information teaching features based
on data. There are three primary strata Layers of convolution, activation, or ReLU, and pooling layer.
 Convolution layer: The convolutional layer places entering photos using a set of customer-
friendly filters, each of which activates a different feature of the image.
 Rectified linear unit (ReLU): By reducing negative qualities to null and keeping positive scores,
the Rectified linear unit (ReLU) speeds up and simplifies training. Because only enabled qualities
are conveyed to the next level, it is sometimes referred to as activation.
 Pooling layer: Pooling increases production efficiency by reducing the number of variables that
the network must understand.
These procedures are repeated over tens or hundreds of layers, learning to discern various properties
with each coating.

Figure 3. 7: Network with many Convolution layers

An example of a multi-layered network is shown in Figure 3.7. Each training image has different
filters applied at different sizes, and the output of each converted image can be utilised as the input
for the next layer.

 Advantages of CNN:
 Convolutional neural networks are so termed because of a data processing phase called
convolution.
  Convolution is beneficial in collecting information characteristics, and even more so in
aggregating and filtering those qualities, hence minimising the number of characteristics (i.e.
pooling).
 This built-in feature reduction is what allows such networks to analyse data with a lot of
features.

3.5.2 RECURRENT NEURAL NETWORK

Recurrent neural network (RNN) extends the capabilities of a typical neural network as shows in
figure 3.8, which can only handle fixed-length data inputs, to accommodate variable-length input
sequences. The RNN processes inputs one by one, using the hidden units' output as additional input
for the obtained element, the layers of RNN is as shown in Figure 3.9. RNNs are more challenging
to train than other types of neural networks because gradients can simply vanish or explode. However,
advances in training and construction have resulted in a range of easier-to-train RNNs. As a result,
RNNs have demonstrated their ability to anticipate the next word in a sentence, speech recognition,
image captioning, language translation, and other time-series prediction tasks. An RNN's hidden units
can keep track of a "state vector" that holds a memory of previous events in the sequence.

Figure 3. 8: Recurrent neural network

Depending on the type of RNN node utilized, the length of this "memory" can be changed. The longer
the memory, the longer the RNN can learn long-term dependencies. RNNs can now handle problems
that require long-term memories thanks to the introduction of long short-term memory (LSTM) units.
A memory cell is a structure in LSTM units that stores information and relates it to itself in the
following time step. New input is added to the memory cell's values, and a forget gate is used to
weight newer and older information differently. The gated recurrent unit (GRU) is another RNN unit
built for long memory. GRUs are comparable to LSTM units, however they have fewer parameters
and are therefore easier to train.

Figure 3. 9: Structure of RNN

3.5.3 COMBINATION OF CNN and RNN

For mesh data, CNN networks have become standard machine learning method (pictures, videos,
etc.). On the basis of CNN model, researchers have developed a range of structural network models
that have proven to be successful in a variety of image correlation issues, including handwritten
digital recognition, natural picture recognition, and so on. AlexNet is a Convolution + Pool structure
neural network that is one of the forerunners of deep learning. Beginning with AlexNet, after
then several CNN models appeared, and ZF-Net began to employ de-convolution to produce the
CNN network intermediate layer feature, which allows people to comprehend working principle and
network architecture more intuitively. In shallow CNN detection images, they discover
characteristics such as edges, colours, and so on, while a deeper CNN layer detects shape properties
of objects to be identified. While the RNN has become an industry standard for sequence data
machine learning (audio, natural language, etc.). The most recent applications include natural
language processing and machine translation. The main research fields using RNN processing
sequence and CNN to process image information are image tagging, target detection, video screen
behavior detection, and so on. Using RNN Layer to Calculate Continuity Features in CNN Layer,
and then connecting these features to final fully connected network for classification prediction.
Figure 3.10 depicts the entire model diagram. The RNN model may extract continuous dependence
features from the intermediate CNN layer's output, and both the CNN and RNN layers can be
extended to many levels.

Figure 3. 10: CNN-RNN structure

3.6 PERFORMANCE METRICS

After the extraction process, analyze the findings using PSNR, MSE, and SSIM, and compare the
two video steganography algorithms using these characteristics.
The fundamental goal of steganography techniques is to hide the secret data inside cover video data,
which means that the quality of cover information will be altered in some way, ranging from little
changes to major distortions. Statistically, several metrics have been used to determine if the distortion
level is acceptable or not.

 PSNR: The Peak Signal-to-Noise Ratio is a normal ratio used to establish the distinction among
transporter and stego information. PSNR is the proportion between the utmost conceivable
 intensity of a flag and the intensity of debasing clamour that influences the constancy of its
representation.
𝐿^2
PSNR = 10 log 𝑀𝑆𝐸 ………..................................................... (2)

Where, L - Maximum intensity it takes 255

PSNR values typically range from 30 to 50 dB, with greater being desirable.

 MSE: By comparing stego and cover files with each of the bytes, the Mean Squared Error is
calculated. The normal squared contrast between the assessed qualities and what is evaluated is
measured by MSE.
1
MSE = 𝐻∗𝑊 ∑ℎ𝑖=0(𝑃(𝑖, 𝑗) − 𝑆(𝑖. 𝑗))2..................................... (3)

 SSIM: The Structural Similarity Index (SSIM) is a perceptive metric that assesses picture quality
degradation as a result of processing such as data compression or data transmission losses.

SSIM (x, y) = (2𝜇𝑥𝜇𝑦 + 𝐶₁)(2𝜎𝑥𝑦 + 𝐶₂ ……………… (4)

( µ𝑥² + µ𝑦² + 𝐶₁)(𝜎𝑥 2 + 𝜎𝑦 2 + 𝐶₂)

To conclude with this chapter gives detailed explanation about the methodology of two video
steganography techniques proposed in this project and also how it was implemented to get expected
outcomes.
CHAPTER 4

SYSTEM REQUIREMENTS
This chapter describes about the hardware and software requirements used in our proposed work and
also the brief explanation about MATLAB software.

4.1 HARDWARE AND SOFTWARE SPECIFICATION

 Hardware:
 System: Pentium IV 2.4 GHz.
 Hard Disk : 40 GB.
 Monitor: 15 VGA Colour.
 Ram: 512 Mb
 Software:
 OS: Windows XP/ Windows 7.
 Software Tool: Matlab 2014a.
 Toolbox: Image processing toolbox.

4.2 INTRODUCTION TO MATLAB

MATLAB is an acronym that stands for MATrix LABoratory. The matrix tools in the LINPACK
(linear system package) and EISPACK (eigen system package) programmes were created in
MATLAB to make them more accessible.

MATLAB is a performance-oriented technical computing language. It's a one-stop shop for

programming, visualisation, and computation. MATLAB is an outstanding teaching and research tool
as a result of these capabilities.

When it comes to tackling technical difficulties, MATLAB has numerous advantagesover traditional
computer languages (e.g., C, FORTRAN). MATLAB is a computer programme that uses an array as
its main data element and does not require dimensioning. The software programme has become a
fundamental tool in most colleges and businesses throughout the world since its initial release in
1984.

It has a set of sophisticated built-in algorithms that can do a variety of calculations. Simple graphics
commands are also included, allowing for quick results viewing. Toolbox packages contain a
collection of specific programmes. Toolboxes span a wide range of areas in applied science and
engineering.

4.3 STARTING MATLAB

To run MATLAB, double-click the MATLAB shortcut icon on your Windows desktop after login.
When you run MATLAB, a new window called the MATLAB desktop appears. Other windows are
contained on the desktop. The following are main tools found on or accessible from the desktop
which is shown in the Table 4.1.

Table 4. 1: Main tools of MATLAB

 Figure 4. 1: The MATLAB workspace's graphical interface

After you first start MATLAB, you'll see a screen similar to the one shown in Figure 4.1. The default
setting of the MATLAB desktop is also shown in this picture. You can rearrange the tools and
documents to suit your requirements.
CHAPTER 5
RESULTS AND DISCUSSION

This chapter describes about results obtained by implementing two video steganography techniques
proposed and also the comparison of results between them.

Using Neural Networks with Hash function and modulo operator in a MATLAB environment, the
steganographic methods provided here effectively hide the secret data in the cover clip. Original cover
clip and embedding data are included in the database. Robustness, imperceptibility, and capacity are
the three basic characteristics of any Steganographic approach.

5.1 SENDER SIDE

First the secret data is provided on the sender side. The secret images and secret text which is being
used for testing the algorithm as shown in Figure 5.1 and 5.2. The secret images and text of different
sizes are embedded into cover frame of cover video and stego video is obtained.

Figure 5. 1: Secret images

Figure 5. 2: Secret text messages

For encryption at sender side our cover media is video, an input video is chosen where the length of
the video is 30 seconds. The selection of input video is as shown in figure 5.3.

Figure 5. 3: Selection of Input Video for Encryption

After selection of input video that video will be converted into frames. The amount of frames
obtained depends on the video which is been selected for this process. Here the video of 30 sec
is converted into 101 frames as shown in figure 5.4.
 Figure 5. 4: Cover video stored as Frames

For the implementation, the secret key that is frame number has to be entered in which secret data
has to be hidden that will be the cover frame as shown in figure 5.5 is being shared personally, so
sender and the receiver have the frame number with them.
 Figure 5. 5: Cover frame

After selection of cover frame, for hiding secret text enter ‘2’ and the text that has to be hidden is
entered by the sender. Then the text is encoded in the selected frame and saved in the encoded video
file.

Figure 5. 6: Embedding secret text 1 using Modulo operator technique

 Figure 5. 7: Embedding secret text 2 using Modulo operator technique

The first “Secret text 1” that consist of 50 characters and “Secret text 2” consists of 150 characters
has been successfully embedded inside the choosen frame using modulo technique and saved in the
encoded video file as shown in figure 5.6 and 5.7.

Using Neural network with hash function technique “Secret text 1” and “Secret text 2” has been
successfully embedded inside the choosen frame and saved in the encoded video file as shown in
figure 5.8 and 5.9.
 Figure 5. 8: Embedding secret text 1 using Neural network using hash function technique

Figure 5. 9: Embedding secret text 2 using Neural network using hash function technique

Enter ‘1' to hide a secret image, then choose the image that the sender needs to hide. Then the image
is encoded in the selected frame and saved in the encoded video file as depicted in figure 5.10.
 Figure 5. 10: Embedding secret image

After embedding secret data using both modulo technique and neural network with hash function
technique in a cover frame, stego frame is obtained as shown in figure 5.11 and 5.12. Then the
parameter values PSNR,SSIM AND MSE is obtained by comparing stego frame and cover frame.
The parameter values will differs for secret image and secret text, as size of the data increases then
there will be increase in the parameter values also. The parameter values is calculated by including
the respective formulas that is discussed in the next section.
Figure 5.11: Stego frame and parameter values obtained using modulo technique
Figure 5. 12: Stego frame and parameter values obtained using neural network technique
5.2 RECEIVER SIDE
For decryption at receiver side, the stego-video created on the sender side is selected as input as
shown in figure 5.13 and used to extract the secret data.

Figure 5. 13: Selection of encoded video

At receiver side, to decrypt the hidden text enter the secret frame number shared by the sender as
shown in figure 5.14.
 If the secret data that has to be decrypted is text, then enter ‘2’.
 If the secret data that has to be decrypted is image, then enter ‘1’.
 Figure 5. 14: Secret key entered by receiver at decryption

After the secret key is entered by the receiver, if we choose to decrypt a secret text then hidden secret
text 1 and secret text 2 will be decrypted and original data is obtained as depicted in the figure 5.15
and figure 5.16.

Figure 5. 15: Decryption of secret text 1

 Figure 5. 16: Decryption of secret text 2

If we choose to decrypt a secret image then hidden secret image will be decrypted and original data
is obtained as depicted in the figure 5.17 at the receiver side.

Figure 5. 17: Decryption of secret image

5.3 COMPARISON OF MODULO AND NEURAL NETWORK USING HASH
FUNCTION METHOD
Image and text were chosen as secret message and same cover-media is used for all messages which
is a video with the resolution 360 X 480 pixels to calculate the MSE, SSIM and the PSNR values.
These values are shown in table 5.1. From this table we can see that the mean square error for the
Modulo algorithm is high when compared with the Neural network using hash function algorithm,
that is the error is more for Modulo. Also we can see that PSNR value for Neural network using hash
function algorithm is high compared to Modulo algorithm. SSIM value is almost same for both
algorithm. As a result, the greater the PSNR number, the more security is provided. This is due to the
fact that it indicates the smallest difference between the original and stego values. The lower the
MSE, improves the proposed algorithm's performance. Hence from table 5.1 neural network using
hash function provides better performance and high security.

Table 5. 1: Comparison of PSNR, MSE and SSIM between Modulo and Neural network using
hash function method

Technique used Secret data PSNR SSIM MSE

entered

Modulo Text 60.4345 0.9997 0.0065

operator

Image 41.1086 0.9775 0.7672

Neural network Text 62.7298 0.9999 0.0037

using hash
function

Image 51.7397 0.9980 0.1268

To conclude with this chapter, it describes about the brief explanation of results obtained by
implementing the methodology of two video steganography techniques proposed and also compared
results of both the techniques.
CHAPTER 6
CONCLUSIONS AND FUTURE SCOPE
Two people can converse privately using video steganography. The proposed method can be used to
securely communicate sensitive data without the risk of being intercepted by a third party. In
steganography, the use of a neural network with a hash function technique provides more security
because it is more difficult for an attacker to find the encrypted pixels. The key benefit of utilizing a
neural network is that decoding the stego-object is extremely difficult, and Hybrid neural networks
handle both kinds of neural networks' shortcomings. For the same message and cover-media, the
Neural network utilizing the hash function algorithm has a higher PSNR value than the Modulo
algorithm.

6.1 APPLICATIONS
1. Military: Secure information transmission is a fundamental goal of good communication in
the military and defense domains. Steganography has long been used by military personnel
for surreptitious communication, as a common transmission channels can attest. Authorized
communication is the foundation of military and defense communication networks.
Additional multiple layered security is offered in the military and defense by applying
numerous encryption algorithms before embedding the sensitive data.
2. Intelligent agencies: The majority of communication in intelligence services is also secret,
and steganography is frequently used for concealment. Steganography has been utilized by
the National Security Agency (NSA) to send secret messages both inside and outside the
agency.
3. Medical: In the sphere of medical sciences, image steganography has been used in diagnostic
scans like as CT, MRI, and others. It's utilized to conceal the patient's sensitive personal
information in the medical imaging itself. Furthermore, a DNA sequence has been employed
to conceal confidential information in medical sciences. This is frequently done to protect the
patient's details and prevent confidential information from being leaked by an unauthorised
person or party.
4. Multimedia: It is frequently used in multimedia applications for copyright information
watermarking. This is done to protect media such as photographs, videos, and other types of
media from unauthorized copying and infringement.
5. Smart IDs: It is now being utilized in smart IDs to hide people's personal information in their
photos.
6. Corporate: Data leakage is the most dangerous aspect of corporate and industry communication.
As a result, steganography-enabled secure communication is always valued for its protection and
authenticity, as risky communication can occasionally result in massive data breaches.

6.2 FUTURE SCOPE

This work can be improved in the future by employing more complicated steganographic and
encryption techniques. AES encryption can be done even before the message is embedded in the
frame. This work can be improved even further by adding some more parameters to reduce noise. In
the future, the frame number might be buried somewhere in the video to make communicating this
information easier.
 REFERENCES

1) Johri, P., Mishra, A., Das, S., and Kumar, A. Survey on steganography methods (text, image,
audio, video, protocol and network steganography). In 2016 3rd International Conference on
Computing for Sustainable Global Development (INDIACom) (March 2016), pp. 2906–2909.
2) Joseph, P., and Vishnukumar, S. A study on steganographic techniques. In 2015 Global
Conference on Communication Technologies (GCCT) (April 2015), pp. 206–210.
3) Eman Abdelfattah, R. J. M., and Elleithy, K. M. Video steganography techniques: Taxonomy,
challenges, and future directions. Systems, Applications and Technology Conference (LISAT),
2017 IEEE Long Island (May 2017).
4) Prashant Johri, A. M., and Das, S. Survey on steganography methods, Computing for Sustainable
Global Development (INDIACom) (February 2016).
5) Sharma, S., Gupta, A., Trivedi, M. C., and Yadav, V. K. Analysis of different text steganography
techniques: A survey. In 2016 Second International Conference on Computational Intelligence
Communication Technology (CICT) (Feb 2016), pp. 130–133.
6) Zaheer, R., Gaur, R. S., and Dixit, V. A literature survey on various approaches of data hiding in
images. In 2017 International Conference on Innovations in Information, Embedded and
Communication Systems (ICIIECS) (March 2017), pp. 1–5.
7) Gupta, H., and Chaturvedi, D. S. Video data hiding through lsb substitution technique.
International Journal Of Engineering And Science 2 (April 2013), 32–39.
8) M.Mary Shanthi Rani, G. M., and Euphrasia, K. Video- steganography based on integer haar
wavelet transforms for secured data transfer. Advances in Intelligent systems and Computing,
Springer series 412, 403–412.
9) Sharma, N., and Batra, D. U. A review on spatial domain technique based on image
steganography. : Computing and Communication Technologies for Smart Nation (IC3TSN)
(2017).
10) Shirali-Shahreza, S. and Shirali-Shahreza M., “Steganography in Textiles”, 4th International
Conference on Information Assurance and Security, pp.56- 61, 2008.
11) Singh, K. M., Singh, L. S., Singh, A. B. and Devi, K. S., ”Hiding Secret Message in Edges of the
Image”, International Conference on Information and Communication Technology (ICICT),
pp.238-241, 2007.
12) Rahate, N. D. and Rothe, P. R., “Data Hiding Technique for Security by using Image
Steganography”, International Conference on Industrial Automation and Computing (ICIAC),pp
 33-36, 2014.
13) Caldwell, J., “Steganography using the technique of orderly changing pixel componenet”,
International Journal of Computer Applications, Vol.58, No.6, 2014.
14) Delahaye, J. P., “Embeddeed Information, Information Hiding”, Scientific American, pp.142-46,
1996.
15) Provos, N. & Honeyman, P., “Hide and Seek: An introduction to steganography”, Security and
Privacy, Vol.1, pp.32-44, 2003.
16) Avcibas, I., Memon, N., and Sankur, B., “Steganalysis Using Image Quality Metrics”, IEEE
Transactions on Image Processing, Vol. 12, No. 2, pp 221- 229, 2003.
17) Sethi, N. and Sharma D., ; “A New Cryptology Approach for Image Encryption”, 2nd IEEE
International Conference on Parallel, Distributed and Grid Computing, pp.905-908, 2012.
18) Johnson, N.F. and Katzenbeisser, S.C., “A survey of steganographic techniques”, International
Journal of Computer Applications, 2004.
19) Wayner, P., “Disappearing cryptography”, 2nd ed. Morgan Kaufmann Publishers, 2002.
20) Karanjit Kaur & Baldip Kaur (2018). “DWT-LSB Approach for Video Steganography using
Artificial Neural Network”. In International Advanced Research Journal in Science, Engineering
and Technology, IARJSET.
21) Anamika Saini, Kamaldeep Joshi, Kirti Sharma & Rainu Nandal (2017). “An Analysis of LSB
Technique in Video Steganography using PSNR and MSE”. In International Journal of Advanced
Research in Computer Science. IJARCS.
22) Ramadhan J. Mstafa and Khaled M. Elleithy & Eman Abdelfattah (2017). “Video Steganography
Techniques: Taxonomy, Challenges and Future Directions. 2017 IEEE.
23) Sushmita MC, Suresh HN & Manikandan J (2017). “An Approach Towards Novel Video
Steganography for Consumer Electronics”. In 2017 IEEE International Conference on Consumer
Electronics- Asia (ICCE-Asia).
24) Hemant Gupta and Dr Setu Chaturvedi,”Video Data Hiding Through LSB Substitution
Technique,” International Journal Of Engineering And Science, vol.2, no. 10 , Issn(e): 2278-
4721, Issn(p):2319-6483, pp. 32-39, April 2013.
25) Mritha Ramalingam and Nor Ashidi Mat Isa, ”VideoSteganography based on Integer Haar
Wavelet Transforms for Secured Data Transfer,” Indian Journal of Science and Technology, vol.
7(7), ISSN (Online) : 0974-5645,pp. 897– 904, July 2014.
26) Sachin Jangid, Somesh Sharma (2017) International Conference on Intelligent Computing and
Control Systems ICICCS, IEEE.
27) S. D. Hu and U. K. Tak, “A Novel Video Steganography based on Non-uniform Rectangular
Partition”, IEEE international conference on computational science and engineering, pp. 57-61,
Aug. 2011.
28) K. Dasgupta, J. K. Mondal and P. Dutta, “Optimized Video Steganography using Genetic
Algorithm (GA)”, International Conference on Computational Intelligence: Modeling,
Techniques and Applications- Elsevier, vol. 10, pp. 131-137, Dec. 2013.
29) S. Bhattacharyya, A. Khan, A. Nandi, A. Dasmalakar, S. Roy and G. Sanyal, “Pixel Mapping
Method (PMM) Based Bit Plane Complexity Segmentation (BPCS) Steganography”, IEEE
World Congress on Information and Communication Technologies (WICT), pp. 36 – 41, 2011.