A Review on Security Techniques in Image Steganography

Abstract

Given the increased popularity of the internet, the exchange of sensitive information leads to concerns about privacy and security. Techniques such as steganography and cryptography have been employed to protect sensitive information. Steganography is one of the promising tools for securely exchanging sensitive information through an unsecured medium. It is a powerful tool for protecting a user's data, wherein the user can hide messages inside other media, such as images, videos, and audios (cover media). Image steganography is the science of concealing secret information inside an image using various techniques. The nature of the embedding process makes the hidden information undetectable to human eyes. The challenges faced by image steganography techniques include achieving high embedding capacity, good imperceptibility, and high security. These criteria are interrelated since enhancing one factor undermines one or more others. This paper provides an overview of existing research related to various techniques and security in image steganography. First, basic information in this domain is presented. Next, various kinds of security techniques used in steganography are explained, such as randomization, encryption, and region-based techniques. This paper covers research published from 2017 to 2022. This review is not exhaustive and aims to explore state-of-the-art techniques applied to enhance security, crucial issues in the domain, and future directions to assist new and current researchers.

Full text

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
361 | P a g e
www.ijacsa.thesai.org
A Review on Security Techniques in Image
Steganography
Sami Ghoul1, Rossilawati Sulaiman2, Zarina Shukur3
Faculty of Information Science and Technology, Universiti Kebangsaan Malaysia UKM Bangi 43600, Malaysia1, 2, 3
Department of Computer Systems Engineering-Faculty of Engineering, University of Zawia, Zawia, Libya1
Abstract—Given the increased popularity of the internet, the
exchange of sensitive information leads to concerns about privacy
and security. Techniques such as steganography and
cryptography have been employed to protect sensitive
information. Steganography is one of the promising tools for
securely exchanging sensitive information through an unsecured
medium. It is a powerful tool for protecting a user’s data,
wherein the user can hide messages inside other media, such as
images, videos, and audios (cover media). Image steganography
is the science of concealing secret information inside an image
using various techniques. The nature of the embedding process
makes the hidden information undetectable to human eyes. The
challenges faced by image steganography techniques include
achieving high embedding capacity, good imperceptibility, and
high security. These criteria are inter-related since enhancing
one factor undermines one or more others. This paper provides
an overview of existing research related to various techniques
and security in image steganography. First, basic information in
this domain is presented. Next, various kinds of security
techniques used in steganography are explained, such as
randomization, encryption, and region-based techniques. This
paper covers research published from 2017 to 2022. This review
is not exhaustive and aims to explore state-of-the-art techniques
applied to enhance security, crucial issues in the domain, and
future directions to assist new and current researchers.
Keywords—Image steganography; data hiding; steganographic
security; randomization; encryption
I. INTRODUCTION
With the evolution of the internet and social networking as
well as a rapid increase in communication facilities, a huge
amount of information is being exchanged every moment.
Security and privacy of exchanged data should be guaranteed,
especially against malicious threats. Cryptography and
steganography are techniques that are being used to achieve
high security [1]. Cryptography is the process of converting
raw information or plaintext to unreadable form called
ciphertext, using an encryption algorithm with a key. The
algorithm and the key utilized by the sender are, in most cases,
used by the receiver for the decryption purpose. Hence, the
original data is unreadable, and confidentiality is maintained.
However, the availability of the ciphertext raises suspicions
and draws the attention of opponents. On the other hand,
steganography seems to be more appropriate and has recently
received much attention, since it does not give rise to any signs
detectable by the human eyes [2]. Steganography is the science
of hiding relatively smaller information in a larger multimedia
cover. Cover media could take the form of text [3], image [4],
audio [5], or video [6]. Image steganography is the process of
concealing secret data within an image that appears normal to
the human eye.
Several reviews on image steganography have been
published in recent years. Study [7] reviews and compares
various deep learning methods in the domain of image
steganography. It categorizes these methods into three types:
traditional, Convolutional Neural Network based, and General
Adversarial Network based methods. It also discusses the
datasets, experiments, and metrics used in the field. However,
traditional steganographic methods are not discussed in this
article. The authors of research [8] provide a comprehensive
overview and evaluation of some of the latest steganographic
methods. This article addresses the challenges of recent deep
learning-based steganographic methods. It also discusses how
to measure the performance of a steganographic technique
using various criteria. Although it evaluates the security of the
techniques vis-à-vis varied advanced attacks and tools,
randomization-related security concepts, such as the chaotic
map function, have not been mentioned or discussed.
The research [9] presents a review and a critical assessment
of recently proposed steganography methods. It describes
various schemes, along with their technical terms, main logics,
as well as strengths and weaknesses in terms of important
measures. This critical assessment is based on the type of cover
object used, the algorithmic domain, and important properties
used as evaluative measures for the steganographic system.
However, this article might be outdated, because many of
contributions have been published since then.
This research article provides an extensive and
comprehensive review of the security principles used in
spatial-domain image steganography. It presents and discusses
various steganographic techniques according to the security
concepts adopted by them, such as randomization, encryption,
and adaptive embedding. Further, it provides an overview of
image steganography, its goals, and traditional steganography
methods along with their features, pros and cons, and
performance evaluation metrics. It also provides analysis,
discussion, and suggestions based on its study of image
steganography security principles.
The remainder of this paper is organized as follows:
Section II presents an overview of image steganography.
Section III demonstrates the basic goals of image
steganography. In Section IV, the search process is defined,
followed by detailing of security techniques in image
steganography in Section V. Section VI contains observations,

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
362 | P a g e
www.ijacsa.thesai.org
discussion, and recommendations. Finally, Section VII
concludes the paper.
II. AN OVERVIEW OF IMAGE STEGANOGRAPHY
The word steganography is of Greek origin and is
constructed from two words: ―steganos‖ and ―graphie‖.
―Stegano‖ means covered and ―graphie‖ means writing.
Steganography is not a new art; it was used in ancient times to
send secret messages by hiding them in certain ways. Fig. 1
summarizes the two ideas used to secure secret information
[10].
Generally, cryptography is divided into symmetric and
asymmetric keys, based on whether it provides authentication
or confidentiality. Likewise, steganography techniques can be
classified into spatial and frequency domain techniques. In the
spatial domain, the secret information is directly embedded in
image pixels; hence, pixel values are modified. On the other
hand, in the frequency or transform domain, the image pixels
are transformed into another domain, and the secret
information is embedded by modifying the coefficient values.
Image steganography is the process of embedding secret
information within a cover image, wherein the embedded
information is not visible to the human eye. The output image,
which contains the secret information, is called a ―stego
image‖ [11]. Fig. 2 depicts a general steganographic system,
wherein the sender optionally compresses or/and encrypts the
secret information and then a certain steganographic algorithm
is utilized to embed the information and produce a stego image.
At the receiver‘s end, the same sequence is followed in reverse
to obtain the embedded information.
A. Traditional Steganography Methods
As mentioned above, steganography techniques are
classified into spatial domain and frequency domain
techniques. In spatial domain techniques, pixel values are
modified directly to incorporate the secret information. These
techniques are famous for attaining high capacity, but they are
not immune to statistical attacks and image manipulations [12].
Examples of spatial domain methods include Least Significant
Bit (LSB) Replacement and its successors, Pixel Value
Differencing (PVD), Pixel Indicator Techniques (PIT), and
Exploiting Modification Direction (EMD) techniques, among
others. In the transform or frequency domain techniques, the
image pixels are first transformed into the frequency domain.
Subsequently, the secret information is hidden by modifying
the coefficient values. Finally, the inverse transform is applied
to obtain the stego image. These techniques resist statistical
attacks but achieve low capacity. The most well-known
transform domain methods are the Discrete Cosine Transform
(DCT), Discrete Fourier Transform (DFT), Discrete Contourlet
Transform (CT), and Discrete Wavelet Transform (DWT)
[12],[13].
Data Security
Techniques
Cryptography Steganography
Symmetric Key Asymmetric Key Spatial Domain Frequency Domain
Fig. 1. Data security classes [7].
Steganographic Algorithm
(Embedding ) Steganographic Algorithm
(Extracting )
Compression
Encryption
Stego Image Stego Image
De-Compression
Decryption
Internet
Cover Image
Secret information Secret information
Fig. 2. General image steganographic system.
LSB Replacement (or simply LSB) is a well-known spatial
domain technique, widely utilized due to its simplicity and
potential to achieve high capacity. In this method, the secret
information is embedded within the least significant bit of the
cover image pixels, in order to minimize distortion. The
resultant stego image is indistinguishable from the original
image. Further, to increase payload capacity, more than one
LSB of a pixel may be used to hide information; however, this
might degrade the visual quality of the stego image [14]. It is
worth mentioning that several enhancements of the original
LSB have already been introduced. LSB Matching (LSBM) is
an enhanced version of the LSB method [15]. Here, if the
embedded bit does not match the cover pixel‘s LSB, then +1 or
-1 values are added randomly to that pixel, so as to avoid the
asymmetry artefacts that are usually introduced by the standard
LSB technique and can be detected by steganalysis techniques
[16]. To improve upon the previous methods, the LSB
Matching Revisited (LSBMR) method was introduced by [17].
It embeds secret information within the LSB, with minimum
changes to the carrier image. It hides two secret bits into two
pixels simultaneously, wherein the first bit is embedded
directly, while the second secret bit value is produced based on
the relationship between those two bits. The objective is to
make the detection of the secret information more difficult,
compared to standard techniques [14], [17].
Another approach towards embedding secret information in
the cover image consists of hiding bits in the edge area where
pixels‘ intensity values change abruptly. Such techniques are
referred to as Edges Based Embedding (EBE) Steganography,
which allows the hiding of large payloads in those particular
edge pixels. Several research studies have been published in
this context, such as [16],[18],[19],[20]. PVD is another
method of hiding binary data, wherein the cover image is
considered to be in the form of non-overlapped blocks of two
pixels each. The difference between the two pixels is calculated
and quantized into several regions that determine the number
of bits of the payload [21]. In the Cyclic Steganographic
Technique (CST), the embedding process is cycled through
color channels of consecutive pixels [22]. The color channel
selected for the current pixel is not the same color channel used
in the previous pixel, or the next pixel. For example, if the LSB
of the red channel is selected for the current pixel, then the
green channel is selected for the next pixel, and the blue
channel for the following pixel. Hence, the same pattern is
preserved for three consecutive pixels. The concept of
randomization along with CST is proposed in [23] wherein
secret information is randomly hidden in the pixels‘ LSB.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
363 | P a g e
www.ijacsa.thesai.org
Fig. 3. Traditional steganographic techniques.
PIT is another variation of LSB techniques and is used to
enhance the security and robustness of classical systems. In this
method, one of the color channels of the pixel is selected as an
indicator for the other two channels that are used for the
embedding process [12]. An improvement is introduced in [24]
takes into consideration the length of the secret message and
uses two LSBs of a particular channel to indicate the presence
of data in other channels. Another variation is introduced in
[25]; it uses three LSBs of one of the pixel‘s channels as an
indicator. EMD is another technique used to enhance security.
In this method, the cover image is partitioned into blocks of n
pixels. These n pixels of the cover image involve secret digits
in a 2n+1-ary notational system. One pixel is eventually altered
by ±1. For a group of n, there are 2n+1 possible digits to be
secretly hidden, since there are 2n possible pixels‘
modifications and one case with no changes [26]. Article
[27]presents a scheme to improve the EMD method called
improved EMD (IEMD). This scheme uses an 8-ary notation
system, wherein the secret digit is embedded into a group of
two pixels. Fig. 3 summarizes the steganographic techniques
explained above.
III. BASIC GOALS OF IMAGE STEGANOGRAPHY
Several considerations must be taken into account while
designing a steganography algorithm: capacity,
imperceptibility, and security. The ultimate goal is to attain
high capacity, better imperceptibility, and high security.
However, these considerations are conversely related, and a
trade-off needs to be made among the criteria mentioned
above. More details are given in the following section
[13],[28],[29].
A. Capacity
It refers to the amount of data in bits that can be embedded
in the cover image. The objective is to hide as many bits as
possible in the cover image, without affecting image
imperceptibility and security. Embedding rate or Bits Per Pixel
(BPP) is another widely-used term, which refers to the total
amount of information relative to the total number of the cover
image pixels.
B. Imperceptibility
When the secret information has been hidden, the resultant
stego image should not create any signs that might be
suspected by human eyes. Hence, the aim is to make the level
of deterioration as low as possible. High imperceptibility
means low capacity and vice versa. Hence, these two concepts
need to be balanced.
C. Security
Security is the main key to secure information sent over an
unsecured network such as the internet. It refers to the ability to
withstand the detectability of hidden secret information in the
cover image as well as the capability of extracting it.
Accordingly, the steganographic algorithm should resist the
attacker‘s attempts to detect and extract the secret information.
As stated before, the aim is to achieve a high embedding
rate, high imperceptibility, and high security. However, this is
a challenging task since these metrics compete against each
other. In other words, focusing on one of them will sacrifice
one or more other metrics. For example, embedding high
payloads will definitely lead to low imperceptibility and the
possibility of low security.
IV. SEARCH PROCESS
A. Materials and Methods
This review study examines the existing methods and
techniques of image steganography security researched
between 2017 and 2022. Only studies that involved spatial
image steganography are considered. This section includes
subsections on data sources, search processes, data selection,
and data extraction.
B. Data Sources
The search and downloading phase has been implemented
intermittently over the period from November 2021 to
February 2022. The primary sources for the research have been
selected from the following libraries:
• Institute of Electrical and Electronics Engineers Xplore
Digital Library (https://ieeexplore.ieee.org/Xplore/
home.jsp.
• Science Direct (https://www.sciencedirect.com/).
• Springer Link (https://link.springer.com/).
• Google Scholar (https://scholar.google.com/)
• Association for Computing Machinery (ACM) Digital
Library (https://dl.acm.org/).
In addition, some other resources have been considered as
well, such as the International Journal of Advanced Computer
Science and Applications (IJACSA).
C. Search Process
The search has focused on image steganography security
and various keyword patterns have been used in this process.
Boolean operators have helped refine the data on keywords for
each research publication. Symbols and Boolean operators
such as ―OR‖ and ―AND‖ have been used to look for the
following keywords:
 ((digital image steganography) OR (security in image
steganography) AND (randomization OR chaotic) AND
(spatial domain).
Steganography
PVD
Transform Domain Spatial Domain
LSB
DCT
DWT
CST
DFT
PIT
EMD
LSBR
LSBM
EBE
LSBMR

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
364 | P a g e
www.ijacsa.thesai.org
 ((chaotic map) AND (image steganography) AND
security)) OR (randomization AND (image
steganography)) AND (spatial domain).
 (((edge detection) OR (region based)) AND ((LSB
image steganography) OR (LSB image
steganography))) AND (spatial domain).
 (cryptography AND (image steganography)) OR
(encryption AND (image steganography)) AND (spatial
domain).
D. Data Selection
We have applied three filtering steps to select the relevant
studies from the search results based on our keywords. The
first step involves specifying the criteria for selecting the
studies. Next, in the second step, the titles and abstracts of the
studies are reviewed according to the research question in the
second step. The third step involves reading the full texts of the
selected studies and extracting the data. The following criteria
have been mostly used for data selection:
 Has the paper been published in the last five years or
so?
 Does the research article mention or discuss any of the
security concepts in the image steganography field?
 Has the research article been included in any of the
reference data sources?
E. Data Extraction
We have examined each preliminary study to identify if it
was related to the security of image steganography. We have
found about 220 papers upon concluding the search in
February 2022. We have selected the relevant ones based on
the criteria mentioned earlier. Finally, 100 related studies have
been identified.
V. SECURITY TECHNIQUES IN IMAGE STEGANOGRAPHY
One of the most challenging steganography aspects is the
security attribute. Here, security refers to the process of making
the hidden secret information undetectable or the embedded
size and locations unguessable. Much research has been
conducted in this domain, which employs different approaches
to improve steganographic security. Fig. 4 depicts a summary
of the findings related to securing image steganography, and
Fig. 5 classifies research articles accordingly.
Encryption is a concept widely utilized to achieve security;
it encrypts the secret information before embedding it. Also,
the entire image may be encrypted as an intermediate step in an
algorithm. Traditional encryption algorithms such as Rivest,
Shamir, and Adleman (RSA), Advanced Encryption Standard
AES, and Triple Data Encryption Standard (3DES) are utilized
to achieve this goal. User-defined techniques are employed as
well. Randomization is another approach towards attaining
security, which embeds secret information in a randomized
way by scattering it over the original image. As part of this
concept, chaotic function, pseudorandom number generator,
user-defined keys, and other techniques can be exploited. In
addition, randomization can be used alone or combined with
encryption techniques to add an extra layer of security. Chaotic
functions are famous for their random behaviour, wherein the
outputs are unpredictable for certain input parameter values.
The output is then exploited in the process of pixel location
selection. Another technique to hide secret information and
enhance security is transforming the cover image into another
bit plane, embedding, and then retransforming the cover image
to the original form. The region-based concept also improves
the overall security, since it breaks the adjacent pixel
correlations. In addition, some other techniques using different
approaches are implemented. In the following section, more
details about the above-mentioned categories have been
presented.
Steganography using
different concepts
Pixel indicator techniques
(PIT)
Region Based
Steganography
Randomization Based
Steganography
Bit-Plane Systems
Encryption
Security Techniques in Image
Steganography
Image Encryption Randomization and
Encryption Steganography Encryption Only
Steganography
Fig. 4. Summary of findings.
Fig. 5. Classification of the existing methods based on security technique.
A. Randomization based Techniques
These techniques rely merely on randomization to achieve
security. In [30], random pixel positions are picked using
Linear Congruential Generator (LCG). Eight bits from the
secret message are embedded using the LSB technique with a
sequence of 3-3-2, which means embedding three bits in the
red channel, three bits in the green channel, and two bits in the
blue channel. Authors in [31] suggest hiding three binary
images within a grayscale image. Binary pixel values are
rearranged using the mean of three random series and are
hidden in a grayscale image employing the concept of Ultra
Unique Numbers (UUN). Random numbers are used for pixel
selection by generating multiple series in [32]. Similarly, [33]

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
365 | P a g e
www.ijacsa.thesai.org
uses Pseudo-Random Number Generator (PRNG) to choose a
pixel for the embedding process. Three PRNGs are generated
with the help of the Skew Tent Map (SKTM) in [34]. They are
employed to scramble secret messages before insertion, choose
an embedding color channel, and select a pixel location with a
particular Red-Green-Blue (RGB) channel.
Authors in [35] proposed an improved 1D chaotic system
model for choosing an embedding pixel location. The chaotic
system utilizes a chaotic Logistic Map (LM) and a sine map
function. A secret message is embedded at random positions
using the Beta Chaotic Map (CM) in [36]. The algorithm in
[37] selects random bits for embedding based on the use of
PRNG. The key seed value and number of embedded bits are
user-defined. Randomization of pixels is done via random
chain codes of the key determined by the user in [38]. These
chains contain a random sequence of bytes based on the
hexadecimal representation of the bytes in the current key
block. This sequence is used to embed the bits of secret
message within the LSB of the pixels in the sub cover image.
The authors of [39] suggest using two secret keys to randomize
one bit of secret message: K1 chooses a channel, whereas K2
places the secret bit within a pixel. The Knight‘s Tour (KT)
algorithm using the LSB technique is suggested by [40]. Here,
the cover image is considered to be the surface of a chessboard,
wherein the knight travels once to each square. In [41], the
secret message is embedded in un-patterned fashion based on
the outcomes of quadratic equations. In addition, combinations
of RGB channels and image partitions are produced, wherein
the Hungarian algorithm is employed to choose the least noisy
combination, that is, the partitions of the cover image and the
secret message.
In [42], standard deviation is utilized to select a richly-
textured block of pixels to hold the secret message. Next, four
Most Significant Bits (MSBs) of three diagonal pixels in the
treated block are selected using SKTM to generate three
correcting bits [Hamming code H (7,4)]. Two of these bits are
XORed with the two secret bits and embedded in the
neighboring pixels. In [43], the embedding position is located
by a key generated utilizing a chaotic LM. Next, these located
positions are further optimized by using two approaches. In the
first approach, Arnold‘s Cat Map (ACM) is applied to add
more randomness by shuffling the pixels. In the second
approach, LM parameters are adjusted using the Genetic and
Bat algorithms to find the best key value for the LM, leading to
minimal changes. In the scheme proposed in [44], two chaotic
maps are considered for pixel selection purposes: 1-D (Tent
map) and 2-D (Baker‘s map). The embedding is done within
the red channel. To minimize distortions, pixels on the edges
are avoided, while the embedding process employs one bit or
two bits of LSB. In the method proposed in [45], the cover
image is scrambled using Power Modulus Scrambling (PMS)
with the aid of the Brownian motion concept. Lighter pixels or
pixels with less intensity exhibit more randomness than the
heavier, darker pixels. The embedding process is carried out
considering certain conditional factors. Finally, the reverse
Brownian-based scrambling is applied to generate the stego
image.
A keyed PRNG is used to scramble pixels, in order to
achieve random permutation, in [46]. The permutation order is
first based on an 8×8 Sudoku puzzle. Next, the embedding is
done using LSB Matching, wherein two bits are embedded in
the red channel, one bit in the green channel, and three bits in
the blue channel. The Cross-coupled chaotic system (using two
chaotic LM) is utilized in [47] to generate a DNA sequence,
which is then added to the secret message‘s DNA sequence
using the ASCII format. The result is then embedded using the
typical LSB method. Table I summarizes the references related
to randomization-based methods.
TABLE I. SUMMARY OF THE MOST OF THE MENTIONED RANDOMIZATION-BASED TECHNIQUES
Ref
Features and Pros
Cons
PSNR (dB)
Evaluation metrics
[30]
 Cover image is 90 angular transformed
 Security is achieved through pixel randomization
using linear congruential generator
 LSB embedding following 3R-3G-2B pattern
 Few experiments
 Embedding does not take into account
the image content
 Not compared to similar techniques
 No steganalysis experiments
 Not robust against statistical and
geometrical attacks
512 x 512
PSNR:
R: 64.0484
G: 64.5621
B: 67.362
PSNR, MSE
[31]
 Able to hide three binary images in the cover image
 Shuffling based security using three random number
series with the help of mathematical relations called
UUN
 Simple implementations
 Embedding does not take into account
the image content
 No steganalysis experiments
 Not robust against structural and
geometrical attacks
 Low PSNR
PSNR: 37.71: 40.46
PSNR, MSE, SSIM,
histogram analysis
[32]
 Binary image is embedded in a grayscale image.
 Randomization embedding using a random number
series.
 Multiple series and reoccurring numbers are
eliminated.
 All image regions are considered
 No steganalysis experiments
 Not robust against structural and
geometrical attacks
Different resolutions
images
Payloads:
(10:100%) 7680:
76800 bits
PSNR: 83.97: 63.45
PSNR, MSE, Bit
error, histogram
[33]
 Embeds 8bits/pixel 3R-3G-2B
 XORing a bit of 5 MSBs with a secret bit and
embedded in the particular LSB.
 Randomization based on PRNG to choose a pixel and
one of 5 MSB
 Simple yet efficient encryption using XOR operation
 Simple implementations
 Security is about inability to retrieve
the message
 All image regions are involved
 No steganalysis experiments
 Not robust against structural and
geometrical attacks

512x512 RGB bmp
images
Payloads
100: 262144 bits.
PSNR:
39.263: 73.798
PSNR, MSE

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
366 | P a g e
www.ijacsa.thesai.org
[34]
 SKTM to generate three PRNGs to: scramble secret
messages, choose an embedding color channel, and
select a pixel location.
 1 & 4 LSBs versions
 Chaotic map exhibits good statistical features
 To make the algorithm more robust against statistical
attacks
 StegExpose detects approximately
half of the embedded information
truly.
 Security is about the secret bits‘
locations.
 Embedding does not take into account
the image content
 Computational complexity
 Many parameter to generate
pseudorandom sequences
RGB 256x256
1-LSB, C=809:9041
Bytes.
PSNR=65.97 :55.49
4-LSB, C= 809:36,167
Bytes
PSNR= 52.621:36.101
COR, HOM, CON,
ENR and CoC.
MSE, PSNR,
MaxErr, L2RAT,
ENT
SSIM, MSSIM,
FSIM
steganalysis:
StegoExpose tool
[35]
 Improved 1D chaotic behaviour (an improved LM
and sine map)
 Improved robustness against statistical attacks.
 Moderate PSNR
 Histogram has many spikes
 Not enough experiments against chi
square test
 Embedding does not take into account
the image content
 Parameters exchange overhead
Color images
128x192, 192x256
256x288, 256x384
Payload: 24,576:
98,304 B
Avg PSNR= 38.209
PSNR, MSE, Image
Fidelity, Entropy
[36]
 Using Beta chaotic map
 Good Visual quality
 Clear Histogram deformity
 Complexity of the Beta map
 Embedding does not take into account
the image content
 Key exchange overhead
USC-SIPI Image
Database.
PSNR: 57.5 – 56.79
PSNR, MSE
[37]
 PRN generator to define embedding locations.
 user-defined Key seed value & number of embedded
bits.
 Not complex
 Variable embedding capacity
 Embedding does not take into account
the image content
 Key exchange overhead
 Few tests
 Non-standard image
 No steganalysis experiments
Message length:
343: 8866 characters
Avg PSNR= 68.49
PSRN
[38]
 Uses chains of a random sequence of indices (codes)
of the bytes in the carrier image.
 Use of the full capacity of the cover image.
 Robustness, and undetectability have been improved
through extracting chains of randomly selected pixels
from the cover image based on a user key
 Not compared to rival techniques.
 Uses non-standard images.
 Uses relatively large Stego secret key
 No steganalysis experiments
Image size= 147456
Payload=18432 Bytes
Image size=111156
Payload= 13894 Bytes
PSNR avg = 51.31
PSRN
[40]
 Security achieved through encryption of secret
message and randomization using KT algorithm
(self-developed algorithm)
 Performance against Chi-square is improved when
using KT
 Not compared to rival techniques
 Image content not considered
 No numerical results
 Stego-key value exchange overhead
Greyscale 512x512
from
USC-SIPI
[41]
 Finds message & cover image combinations with
minimum changes
 Randomization through using an un-patterned
quadratic embedding sequence with unbounded i/p
parameters.
 An artificially created assignment problem with an
optimized solution of the Hungarian algorithm
 Security is related to the embedding
locations
 Image content not considered
 Stego-key value exchange overhead
 No steganalysis experiments
 Complexity high
Color image: 256x384
Payload= 23KB.
Avg PSNR =52.739
PSNR
[43]
 Randomization of embedding location with
optimizations using Chaotic LM to achieve highest
PSNR possible
 Uses LSB replacement and LSB matching
 Optimization using Arnold‘s Cat map and adjustment
of LM parameters using Genetic and Bat algorithm
 High complexity
 No steganalysis tests
 Key exchange overhead
 Image structure not considered
256x256 grayscale.
Embedding rate:
20:100%
2LSB: PSNR
Org= 47.52
Optimized=48.44
SM2LSB PSNR,
Org= 44.53
Optimized=45.22
PSNR
[45]
 Randomization key generated using Brownian
motion
 Nonlinear Brownian motion adds more security
 High capacity
 Image contents not considered
 Complex
 Payload size not mentioned in the
steganalysis
 Key exchange overhead
128 × 128, 256 × 256,
512 × 512
Avg. PSNR= 48.467
without Brownian
Avg. PSNR with
Brownian = 48.45417
MSE, PSNR, SSIM
entropy, Laplacian
MSE Error
(LMSE), Mean, S.
deviation, Kurtosis,
Skewness, KL
Divergence, and
NCC
[47]
 Randomization is achieved by combining two chaotic
LMs to generate PRNG.
 LM is utilized to generate a DNA sequence
 The sequence is added to the secret message's DNA
sequence using the ASCII format.
 The result is LSB embedded.
 No tabulated experiments
 Image regions not considered
 Key exchange overhead
Payload:
37-character (296 bits)
PSNR
99 dB
PSNR, Coefficient
Correlation Test,
Entropy

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
367 | P a g e
www.ijacsa.thesai.org
B. Encryption
In the following sections, steganographic techniques will be
presented based on cryptography only or combined with
randomization. In addition, some image encryption techniques
will be presented, which can be utilized to encrypt the secret
image and be used during the steganographic process.
1) Encryption based steganography: The technique in
[48] proposes to encrypt the secret message using an XOR
operation with a user-defined key. Next, a 4-bit shifting
operation is applied to the encrypted text message in order to
form the secret message to be hidden. The secret data is
eventually embedded within the cover image using the
standard LSB method. As per the technique suggested in [49],
the letters of the secret message are initially transposed as an
encryption process. Next, the secret message bit is XORed
with the MSB of the image pixel based on a particular key
hidden in the LSB of the cover image. In the method presented
by [50], data hiding is performed using the typical LSB
concept. The secret message is encrypted using the AES
encryption algorithm with a 128-bit key in Cipher-Block
Chaining (CBC) mode. Next, the order of pixel selection for
embedding is obtained by utilizing a combination of the image
attributes such as type and image resolution. The secret
message in [51] is encrypted using XOR operation through a
key generated via a circular bit shift operation having varying
block sizes. Next, the encrypted message is embedded in RGB
channels using the LSB algorithm in various ways for each
channel. This process uses a 2-3-4 paradigm, wherein the
insertion is done sequentially for the red channel, using raster
scan pattern from right-left for the green channel, and using
top-bottom raster scan for the blue channel.
The input RGB image in [52] is scrambled using a hyper-
chaotic map to produce a permuted encrypted version of the
cover image. The encrypted image is converted to YCBCR
color space, and the luminance channel (Y) is then divided into
8×8 non-overlapping blocks to apply DCT and quantization on
each block. Finally, Huffman coding is applied to the secret
message and embedded in the left MSB. In the technique
suggested by [53], the secret message is encrypted using a
symmetric key cipher. The encrypted message is XORed with
selected bits from the cover image for obtaining a higher-order
pixel to add more confusion to the stego image. A block-wise
inversion technique is applied to minimize changes during
embedding by inserting them to an LSB or inverting them. The
stego key consists of a symmetric encryption key and the
encoding key, which contains parameter settings such as the
number of blocks, starting block, start pixel offset, and block
selection rule. In [54], the author introduces a secure
steganography scheme, which encrypts the secret information
using the permutations concept. Two chaotic map functions are
utilized to obtain a sequence that is XORed with secret bits.
The cover image is JPEG compressed, and DCT coefficients
are modified and adaptively selected to hide the secret bits
using a histogram modification-based data hiding scheme. In
[55], the cover image is flipped by 180°. Next, the blue channel
is divided into four sub-blocks, each of which is shuffled. The
difference between the red pixels and the secret message is
calculated and encrypted using a Multi-Level Encryption
Algorithm (MLEA), which includes XOR, permutation, and
shifting operations. The LSB embedding is done within the
blue shuffled pixels. Subsequently, the sub-images are
reshuffled and eventually combined using the red and green
channels and re-flipped. Table II summarizes the encryption-
based method in image steganography.
2) Randomization and encryption based steganographic
techniques: The study [56] presents a scheme to improve the
security of LSB steganography. In this method, the secret
message is encrypted using a One-Time Pad (OTP)
encryption. Subsequently, randomization is achieved by
columnar transposition and RGB color plane scattering
technique. Also, the technique in [13] uses OTP to encrypt the
secret message, but it employs PRNG to select a pixel location
for LSB embedding. In [57], the AES algorithm is applied to
encrypt the secret message, which is then embedded using the
LSB technique at a location randomly determined by the LCG
method. The AES algorithm is also used in [58] to encrypt the
secret message before dividing it into blocks. The cover image
is segmented using a technique called a Non-Uniform Block
Adaptive Segmentation on Image (NUBASI) algorithm.
Finally, PRNG is used to randomly choose a message block to
be embedded into an image segment. In fact, there are 32
predefined pattern orders of segments that can be selected at
random. In [59], the secret message is encrypted and
compressed by employing Vigenère Cipher and Huffman
Coding, respectively. Next, the image is segmented into
blocks in order to apply the KT algorithm to make groups of
blocks. An arbitrary function is employed to select which
blocks and groups can be used to conceal a specific pixel in the
group randomly. Authors of [60] suggest encoding the secret
message using bitwise XOR operations between adjacent
pixels. Next, a local user selection is applied to find a particular
position among the four least significant bits to hide a secret
bit. The technique presented in [61] is based on Modified Least
Significant Bit (MDLSB) to embed data in the cover image. It
uses two layers of randomization: segment selection and pixel
selection based on a user key. Further, a lossless compression
algorithm, the DEFLATE algorithm, is applied to overcome
the issue of embedding size in the subsequent layer. In
addition, the AES encryption algorithm is utilized to encrypt
the secret message in the next layer. The Artificial Bee Colony
(ABC) algorithm is employed to reduce the noise caused by
embedded information.
A randomized key based technique is proposed in [62].
First, the cover image is compressed, and then a secret key
with the same length as the secret message is generated. Next,
the secret message bits are XORed using the generated key.
The resultant bits are embedded in the locations specified by
the key. Finally, the image is encrypted using the 3DES
algorithm. Authors in [63] proposed a secure digital image
steganography scheme. In this approach, the secret message is
initially compressed using the Run Length Encoding (RLE)
algorithm, which is then encrypted using the Bernoulli map
and the inverting bit map technique. During the embedding
stage, the embedding location is selected by using KT
algorithm and Henon Map (HM) function. The cover image is

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
368 | P a g e
www.ijacsa.thesai.org
divided into 8x8 blocks in order to apply the KT for selecting a
block. HM function is employed to choose a pixel location
among the 64 pixels in a particular block. The selected pixel is
decomposed through Fibonacci, so as to embed the secret bit
using a New Stego Key Adaptive LSB (NSKA-LSB), which
maintains good imperceptibility.
In the technique proposed by [64], the location and the
order of the image pixels are randomly chosen for embedding
using a chaotic PRNG. The Pseudo-Random Number (PRN) is
generated by exploiting the Duffing map and Circle map
functions. In addition, the secret message is encrypted by
XORing it with PRN before initiating the embedding process.
In [65], the authors present a secure method that starts with
encryption of the secret message using the RSA algorithm in
combination with the Diffie-Hellman (DH) key exchange
algorithm. Next, the encrypted message is compressed using
RLE. Finally, via the Direct Sequence Spread Spectrum
(DSSS) technique, PRN is used to select random pixels and to
embed the secret message through 1-bit LSB and 2-bit LSB.
The research [66] suggests a new method to enhance
steganographic security. In this technique, the secret message
written in Turkish text is encoded and compressed using the
Huffman coding. The secret message is encrypted using the
XOR operation and an OTP key with an equal-length payload.
The key is generated using the super Mandelbrot sets and with
the help of the LM. Subsequently, the LSB plane is analyzed
morphologically to avoid low entropy pixel locations. Finally,
LSB hiding is applied with the help of the chaotic LM in order
to randomly choose a pixel location.
TABLE II. SUMMARY OF ENCRYPTION-BASED STEGANOGRAPHY TECHNIQUES
Ref
Features and Pros
Cons
PSNR (dB)
Evaluation metrics
[48]
 XOR Encryption with bit-shifting of the secret
message.
 Three RGB pixels used to hide an 8-bit data
 LSB based
 Simple encryption operation
 Image structure not considered
 Steganalysis evaluation is missing
 Not robust against statistical attacks
 Same channel sequence followed
RGB 512x512
87373 bytes
Avg PSNR= 51.637
PSNR, SSIM
[49]
 Transposition of the secret message and XOR with
MSB of the pixels then embedded in LSB
 Less computation and time complexities comparing to
standard encryption techniques
 Steganalysis evaluation is missing
 Secret key exchange
 Not robust against statistical attacks
 Limited capacity
256x256 grayscale
image
Payload: 1: 4 KB.
PSNR= 63.236 : 57.132
MSE, PSNR,
Histogram
[50]
 AES-CBC Encryption of the secret message
 Embedding order of RGB depend on: file type &
resolution
 LSB based
 Steganalysis evaluation is missing
 Not robust against statistical attacks
 Limited capacity
 Secret key exchange overhead
512x512 color images
bpp=1
Avg. PSNR=51.196
RMSE, PSNR, MSE
[51]
 XOR encryption of the secret message with a circular
bit shift generated key
 varying block size
 LSB embedding in the 2-2-4 LSB‘s of the RGB
channels
 Less complexity due to use of XOR encryption
 All image regions considered
 Steganalysis evaluation is missing
 Secret key exchange
 Not robust against statistical and
geometrical attacks
128×128, 512×512,
800×600
Payload: 16 KB: 480
KB
bpp = 8
Avg. PSNR =64.85
MSE, PSNR,
Entropy
[52]
 Encryption by shuffling image rows & columns with a
help of hyper chaotic map
 The Y channel of YCbCr version is DCT quantized
 Embedding the Huffman of secret message into the
MSB
 Huffman coding increases capacity & security.
 Robust against geometric attacks such as cropping
attack, rotation attack, scaling attack
 Steganalysis evaluation is missing
 Parameters exchange overhead
 Tested against low payloads
 Computation complexity
512×512 & 256×256
standard color
256x256 medical
images
Payloads: 100: 2050
Bytes
PSNR=77.16: 53.68
PSNR, MSE, BER,
SSIM, correlation,
Quality Score
prediction, Mean
value, Standard
Deviation,
Entropy, Gradient
[53]
 Encryption followed by XOR encoding of the message
with randomly selected higher-order pixel.
 Resultant bit alterations minimized using a block-wise
inversion.
 Robust against chi-square, RS analysis, and sample
pair test for the majority of stego images
 Use public key to exchange stego-key
 Many parameters
 All image contents considered
 Clear Histogram spikes
 Low payloads
256×256 Greyscale
bpp=0.25: 0.9
PSNR: 57.475:51.629
MSE, PSNR, NCC,
SSIM, Histogram
based analysis
(PDH), chi-square,
RS groups analysis,
sample pair test
[54]
 Hyper-chaotic system to permute the secret message
and XORing it with the chaotic sequence
 DCT coefficients modified using histogram
modification-based data hiding scheme for high
capacity
 Low distortion
 Steganalysis evaluation is missing
 Parameters exchange overhead
 Computation complexity
 Moderate embedding rate
USC-SIPI&UCID
database
512x512
bpp = 0.2671
PSNR= 36.05: 38.93
bpp=0.08:0.18
PSNR=38.322: 41.695
PSNR, MSE, SSIM,
average embedding
rate ER, Information
entropy, Correlation
coefficients
[55]
 Flipping of the cover image and blue channel is
divided and shuffled.
 Multi-level encryption algorithm is applied to encrypt
intermediate values
 The result is embedded in LSB of the blue, reshuffle,
and combined with red & green.
 Image structure not considered
 Steganalysis evaluation is missing
 Parameters exchange overhead
 Computation complexity

512x512 color images
from USC-SIPI
Payloads: 2:8 KB
Avg PSNR= 65.9225
For 256×256
Payloads: 2:8 KB
Avg PSNR= 71.0515
MSE, RMSE,
PSNR,
MAE, NCC and
SSIM

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
369 | P a g e
www.ijacsa.thesai.org
In the scheme proposed by [67], the Bernoulli chaotic map
function is utilized in three stages. It begins with encryption of
the secret message by XORing it using a random sequence
generated by the chaotic function. Then, one of the RGB
channels is selected randomly using the random sequence. In
addition, the secret message bits are randomly embedded in the
rows and columns of the cover image. Embedding is done in
the fourth least significant bit of the pixel. In [68], the secret
data is transformed into binary form and then its CRC-32
checksum is generated. The data and its CRC are Gzip-
compressed, followed by the AES encryption that is eventually
appended to the header information. The last phase consists of
information insertion using the Fisher-Yates Shuffle algorithm.
This insertion enables selection of a pixel location for the
embedding process, wherein all LSB channels are employed.
A method is suggested in [69] which starts by finding the
best place to hide secret data using the LSB technique. Then,
the secret information is encrypted using the 3DES algorithm;
here, the RSA algorithm is used for the secret key exchange.
The last stage is randomization, which is realized using a built-
in randomization function based on the modulo Operation
(mod) function with a secret seed. The secret seed is exchanged
using the RSA. In the three-layered security suggested by [70],
a small size fragile watermark is appended to the secret
information to make it tamper-proof. The data is scrambled
before the embedding process and partitioned into three vectors
of varying length. The largest vector is embedded in the lowest
order bit plane to minimize distortion that may arise after
embedding. Before insertion, the host image is encrypted using
the scrambling notion based on a Pseudo-Random Address
Vector generated for image encryption, which is called PAVE.
The concept of Pseudo Noise (PN) sequence generator is
employed to generate the PAVE. The encrypted image is then
partitioned. The embedding location is selected randomly by
generating a random vector called Pseudo-Random Address
Vector for Data Hiding (PAVH). Once the embedding is
accomplished, the host image is decrypted to obtain the final
stego image.
In [71], the first step towards achieving security comprises
encryption of the secret image using RSA, which produces a
16-bit pixel cipher. The resultant cipher is rearranged to obtain
a binary image, which is subsequently scrambled using a
randomization concept relying on 2-D ACM transformation.
The secret image is then embedded within the cover image
using an inverted 2-bit LSB steganography wherein two bits
are embedded. Bit inversion is based on the 3rd and 4th bits
that are utilized as quantifiers. In [72], a chaotic LM and a
support image are combined to generate a random binary
sequence that is XORed with the secret message to get an
encrypted version of the message. LSB embedding is done at
positions determined randomly through random sequencing.
The support image is pre-processed and utilized during the
embedding stages to help resist steganalysis. In the method
proposed by [73], the sequential color cycle is combined with
pattern-based image steganography. The data to be hidden is
encrypted using the AES algorithm. The cover image is
divided into blocks and sub-blocks. Finally, a sub-block is
chosen based on a predefined bow-tie shape pattern. The
embedding operation is accomplished sequentially for the LSB
of the RGB channels.
An algorithm to enhance security using randomization and
multiple encryptions of secret images is presented in [74]. This
technique embeds three secret binary images inside an RGB
cover image. First, the three-color planes are separated. The red
channel matrix is further separated into even and odd matrices.
Next, the bits of the first secret image are embedded in the LSB
of the odd matrix and the second LSB of the even matrix
containing red pixels. Further, the bits of the second and third
secret images are encrypted by XORing them with the first
secret image. Finally, the two encrypted images are inserted in
the LSB of the green and the blue channels, respectively. A
pixel locator sequence-based technique was suggested by [75]
to enhance LSB steganography security. This scheme starts by
enciphering the secret information using the AES algorithm.
The encrypted information is randomly embedded within the
LSB bits of the image pixels with the help of a pixel locator
sequence. Modern Fisher-Yates shuffle is utilized to generate
the random sequence. The pixel locator sequence is also
encrypted and appended to the image to form the final version
of the stego image.
To secure communication, [76] incorporates randomization
in combination with encryption. Confidential data is first
encrypted using the Blowfish algorithm. Next, the LCG
algorithm is used to generate random numbers that are in turn
used to select pixels for LSB embedding. The number of bits to
be inserted varies depending upon the pixel‘s intensity. Chaos
based steganography is presented in [77]. In the first step, the
secret information is encrypted using the AES algorithm.
Following this, chaotic LM and ACM are employed to
generate random values. These values are used to choose the
positions of pixels on the cover image within which the secret
bits are embedded. The first and second bit planes are used to
uniformly distribute bits of the message. Three variations of
random values can be constructed using the two
aforementioned chaotic maps. Table III summarizes the
references related to randomization and encryption-based
steganography methods.
Image encryption: Various image encryption techniques are
explained in this section. In [78], image encryption based on
the confusion process is followed. The RGB image is
transformed into a square grayscale image in the first step.
Next, ACM is applied to scramble image rows and columns,
followed by the HM algorithm for further scrambling to obtain
the final cipher image. Authors of [79] suggest enciphering
digital images by incorporating the confusion and diffusion
concepts. First, the 256x256 RGB image is divided into four
quadrants. Each of these quadrants is subsequently divided into
four sub-quadrants that are rotated 900 anti-clockwise to form
64 sub-blocks in total. Then, modified zigzag transformation is
applied to each channel to break the association with the
adjacent pixels. Up to this step, confusion is fulfilled. An
Enhanced Logistic Map (ELM) is used to generate
intermediate encryption keys that pick specific pixel values to
guarantee diffusion. The final key ‗K‘ is generated based on
the chosen values from the image and external user key, which
are then XORed with RGB channels produced earlier after the
zigzag step.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
370 | P a g e
www.ijacsa.thesai.org
TABLE III. SUMMARY OF RANDOMIZATION AND ENCRYPTION BASED STEGANOGRAPHY TECHNIQUES
Ref
Features and Pros
Cons
PSNR (dB)
Evaluations Metrics
[13]
 Simple encryption and reduced computation
using XOR encryption
 OTP is randomly generated
 PRNG used to randomize the encrypted secret
message into the cover image
 Simple implementation using LSB Substitution-
based
 Steganalysis evaluation is missing
 Image structure not considered
 Image size not mentioned and payload
 Not robust against statistical steganalysis
 Channels used in the same order
 Parameters exchange overhead
RGB images
bpp=3
Avg PSNR= 83.27
MSE, PSNR, SSIM
[56]
 Simple encryption using OTP encryption of the
secret message
 Randomization is achieved by columnar
transposition and RGB color plane scattering
technique.
 Simple implementation
 Reduced computation and time complexities
compared to conventional encryption
 Steganalysis evaluation is missing
 Image structure not considered
 Image size not mentioned and payload
RGB images
Avg PSNR=58.47
MSE, PSNR, SSIM
[57]
 The secret message is AES encrypted
 Pixels are randomly chosen using LCG method
 Standard LSB embedding
 All RGB LSBs are utilized
 Key exchange overhead
 AES complexity
 Steganalysis evaluation is missing
 Image structure not considered
RGB images. Size of:
607x695 - 1,293×1,480
Payload = 8,230: 76,700
bytes
PSNR=59.29:71
MSE, PSNR
[58]
 Three-layer securities through Encryption &
randomization
 AES and PRNG are used for message
encryption and message block selection
 Uses NUBASI algorithm for the cover image
 Secret key exchange overhead
 Complex
 Steganalysis evaluation is missing
 Image structure not considered
512x512 RGB
Payload= 2.5 KB
Avg. PSNR = 63.755
PSNR
[59]
 Strength against electronic attacks by
encryption, compressions, and randomization.
 Randomization using KT algorithm and
arbitrary function
 Robust against Chi-square attack
 Exploiting Modification Direction embedding
based technique
 Encryption key, arbitrary function key,
and Huffman table exchange overhead
 Complex
 Payload in the steganalysis not
mentioned
 Not tested against other attacks
 Image structure not considered
512x512 images from
USC-SIPI
bpp = 0.5: 1.6
PSNR = 60.84: 55.69
PSNR, MSE and
SSIM
[60]
 XOR encoding of the secret message
 User selection to hide a secret bit in one of the 4
LSBs
 Modified LSB
 Simple implementation
 Multiple cover images possibly needed
 Steganalysis evaluation is missing
 Image structure not considered
 Not compared with similar techniques
256x256 grey scale
payload: 128x128 image
Avg. PSNR = 43.455
MSE, PSNR
[61]
 To enhance security, use of multi-stage
randomization, a lossless compression algorithm
(DEFLATE algorithm), and AES encryption
 ABC algorithm to reduce embedding noise
 Tested using ROC curves of the WFLogSv
steganalyser
 AES key exchange overhead
 Not simple
 Time Complexity
 Image structure not considered
 Other Steganalysis techniques missing
Images from UCID
database
150×150 - 1080×1024
PSNR without ABC=
48.1:58.2
with ABC: enhance by
magnitude of 3:6
PSNR, SSIM
Euclidean norm
testing
ROC curves of the
WFLogSv
steganalyser
[62]
 Security is achieved by Compression,
randomization, and encryption
 The image is compressed and XORed with a
generated key
 A key based randomization to hide the secret
message
 Encryption of the image using 3DES
 Steganalysis evaluation is missing
 Image structure not considered
 Few tests
 Not compared with similar techniques
 Not typical steganography, encryption
alike
Color images: 256x256,
250x360 pixels
Avg PSNR: 53.51
MSE, PSNR, and NC
[63]
 Utilizes compression, encryption (Bernoulli
map), randomization using KT and HM.
 Fibonacci decomposition with the help of
NSKA-LSB
 Maintains good imperceptibility.
 Complex due to many stages
 Steganalysis evaluation is missing
 Image structure not considered
 Few tests
 Not compared with similar techniques
512 x 512 gray images
From USC-SIPI
Embedding rate = 6.25,
12.5, 18.75 %,
Avg, PSNR= 61.14,
66.58, 72.29
PSNR, SSIM, and
BER

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
371 | P a g e
www.ijacsa.thesai.org
[64]
 Randomization with help of Duffing map and
Circle map
 XOR Encryption of the secret message
 Chi Square test passed with payload of 4000
bits
 Good imperceptibility
 LSB replacement based
 Only Chi-square steganalysis is used for
evaluation with low payload
 Image structure not considered
 Key exchange overhead
Non-standard color
images:
256×256, 512×512
Payload: 800:16,000 bits
PSNR:
256x256 = 63.0: 76.56
512x512 = 69.1: 82.73
Randomness tests,
MSE, PSNR and
SSIM
The Average
Difference, Laplacian
MSE, NAE, IQI
[65]
 Secret message encryption and compression
 Randomization to select random pixels to
embed the secret message
 Direct sequence spread spectrum technique is
used in LSB-1 and LSB-2
 Steganalysis evaluation is missing
 Image structure not considered
 Few tests
 Not compared with similar techniques
 Diffie-Hellman to exchange keys
jpg images
payload:
50: 1000 characters
PSNR
58.528: 71.596
PSNR, MSE
[66]
 Regional adaptive with randomization
embedding.
 Huffman compression and OTP-XOR
encryption of Turkish secret message
 OTP generated using Mandelbrot sets with the
LM.
 Only high entropy regions are considered
 Only Chi-square steganalysis is used for
evaluation
 Huffman table and parameter exchange
overhead
 Relatively complicated
 Only LSB plane is analysed and
considered
NASA&SIPI grayscale
images
512x512,
Payload: 1: 93.5 KB
PSNR= 61.9: 51.15
PSNR, BPP, SSIM
UNIVERSAL
IMAGE QUALITY
INDEX (UIQI)
[67]
 XOR encryption of the secret message with a
random sequence of Bernoulli chaotic map
function
 Randomization of channel and pixel selection
 Embedding is done in one of the fourth LSB of
the pixel.
 Steganalysis evaluation is missing
 Not robust due to Image structure not
considered
 Low payloads used
RGB: 256x256, 720×576
Payload:
44: 500 characters
PSNR: 52.12: 67.82
Statistical Metrics.
MSE, PSNR, Max
Absolute Squared
Deviation, Ratio of
the squared norm and
CC
[68]
 First security tier is the secret message
compression and AES encryption
 Pixels randomization using Fisher-Yates Shuffle
algorithm
 Resistant to the Chi-square
 Integrity check guaranteed using message CRC
 High capacity with good imperceptibility
 Only used Chi-square steganalysis
 Image structure not considered
 Key exchange overhead
 Relatively complicated
512x512 RGB
Payload: 1 : 256 KB
PSNR: 40.083: 63.862
Max payload 315392
Bytes (Gzip)
MSE, PSNR, NCC,
Average Difference,
Maximum Difference,
Laplacian MSE and
Normalized Absolute
Error
[69]
 3DES encryption of the secret message,
 Randomization using mod function with secret
seed.
 RSA for key exchange.
 LSB embedding
 Key exchange overhead
 Image structure not considered
 Weak randomization as its not chaotic
 Steganalysis evaluation is missing
Payload:
12: 48 KB
PSNR
53.3150: 52.8734
PSNR, SSIM,
retrieval bit, error rate
(RBER)
[70]
 Pseudorandom vectors used for image
encryption, scrambling and secret message
encryption
 Embedding is Randomly achieved at
Intermediate Significant Bit (ISB)
 High security due to encryption and
randomizations
 Complex
 Master key exchange overhead
 Image structure not considered
 Steganalysis evaluation is missing
 Moderate PSNR
512×512 standard images
bpp=2
Avg PSNR= 39.11
PSNR, MSE,
Normalized Absolute
Error (NAE) and
NCC.
[71]
 RSA Encryption of the secret image
 Randomization using 2-D ACM
 Embedding using an inverted 2-bit LSB
steganography
 Embedded in one channel (blue)
 High capacity
 Image structure not considered
 Steganalysis evaluation is missing
 Time complexity
512x512 images
Secret messages: RGB
64x64 & Grayscale
128x128
Payload: 24,576:32,768
KB
PSNR=57.25: 52.5
PSNR, MSE
[72]
 XOR encryption of secret information with help
of LM.
 Embedding LSB plane randomized
 Support image processed for information
extraction
 Imperceptibility of the first image is ideal
 Image structure not considered
 Steganalysis evaluation is missing
 Extra overhead due to using two images
 Key exchange overhead
128x128 pixels
1st Primary image= 50%
data
2nd Support image=50%
data
PSNR of 1st image =inf
PSNR of 2nd image= 53.24
PSNR, MSE
[73]
 AES encryption of the secret message
 Randomization based on Bow-tie shape pattern
 LSB Embedding using SCC
 Stego image is further encrypted
 Image structure not considered
 Steganalysis evaluation is missing
 Key exchange overhead
 Few tests
RGB 256x256, 512x512
PSNR= 39%, 49%
PSNR

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
372 | P a g e
www.ijacsa.thesai.org
[74]
 The secret images separated into multiple parts
 Based on RGB planes with several
combinations of separated red channel with
multiple XOR encryption of the secret and the
green and blue LSBs
 Randomly embedding in LSB of the green and
the blue channels
 High complexity due to using 3 secret
images
 Image structure not considered
 Steganalysis evaluation is missing
 Key exchange overhead
 Few tests
Cover images jpg
255x255x3
Secret images: .png
LSB & LSB1
PSNR: 49.248
LSB1&LSB2
PSNR:44.999
PSNR, MSE
[75]
 AES encryption of the message
 Randomization of pixels using Modern Fisher-
Yates Shuffle
 Embedding using LSB substitution
 Pixels sequence is encrypted and sent along
with image
 Image structure not considered
 Only tested against Chi-square
steganalysis
 Pixel Locator Sequence exchange
overhead
 Not compared to similar techniques
RGB images
225×225, 512×512
Avg. PSNR =46.148
PSNR, MSE, Chi
square
[76]
 Encrypting the secret message using Blowfish
algorithm
 Randomization achieved by LCG algorithm
 Number of embedded bits is intensity based (2-6
bits)
 Image structure not considered
 Only used Chi-square steganalysis
 Not compared to similar techniques
 Payload capacity is not mentioned in Chi
square attack.
256×256, USC-SIPI
image database
bpp: 2.00: 2.95
Avg PSNR = 47.286
MSE RMSE PSNR
Chi-Square Attack
[77]
 The secret message is AES encrypted and the
embedding positions are randomized using LM
and Cat maps.
 1st & 2nd bit planes are utilized
 Using accurate PSNR (weighted PSNR)
 Image structure not considered
 Non-standard steganalysis techniques
 Not compared to similar techniques
 parameters exchange overhead.
Medical images
256 × 256, 512×512
Payload 0.5 bpp.
Avg PSNR= 50.89: 50.73
Avg wPSNR= 60.21:
69.84
PSNR, wPSNR,
MSE, SSIM, Cross-
correlation
Coefficient, Entropy
and Key sensitivity.
In [80], a color image of H×W size is decomposed into
three images consisting of its RGB components at the bit-level.
These images are then vertically combined to create one bit-
level image (3H rows x 8W columns). Based on Chaotic
SKTM, permutation is carried out at a bit level of the image.
The initial values of the chaotic function are concluded based
on a 128-bit key value, which is constructed from a
combination of external user key and information extracted
from the plain image. Eventually, the function is iterated
3Hx8W times, and the results are recorded and inverted to be
used for permuting the image bits accordingly. A lightweight
encryption scheme is presented in [81]. The raw image is
converted to a 1D array and permuted using a random
sequence produced by the LM. This sequence is transformed
into a DNA sequence. Also, the LM is used to generate another
random sequence that is likewise used to obtain another DNA
sequence. The two sequences are added using DNA
computation rules. In the final step, each pixel is XORed with
its preceding pixel. In [82], the authors propose a compression
and encryption method based on hyper-chaotic function, 2D
compressed sensing, and DNA encoding. The aim of using a
hyper-chaotic function instead of the 1D chaotic function is to
increase the system key space, thereby improving system
complexity and security. The cover gray image is compressed
using the 2D compressed sensing technique. Fractional-order
Chen hyper-chaotic system is employed to control DNA
encoding and operation. The initial values of the Fractional-
Order hyper-chaotic system are generated using the hash value
of the original image. The DNA encoded image is enciphered
using Arnold transformation to produce the encrypted image.
In [83] authors propose an encryption scheme that uses
intertwining and ELMs. The primary point is to de-associate
neighboring pixels by shuffling them. Next, the image is
partitioned into blocks, and permutation-substitution operations
are performed using an intertwining LM and ELM along with
the cosine-based transformation. In addition, the image is
rotated 90° anticlockwise. Finally, random order substitution is
achieved by utilizing the random sequence generated using a
chaotic function. A novel high speed image encryption scheme
based on 1-D cosine fractional chaotic map and image
encryption scheme (DCF–IES) is proposed in [84], which is
based on a new real 1-Dimensional Cosine Fractional (1-DCF)
chaotic map. To increase encryption speed, permutation
operation is excluded from the architecture design. Despite the
absence of the permutation process, a substitution process with
a high sensitivity to a plain image guarantees a high level of
security. The chaotic function is first utilized to generate two
sequences that act as system keys, wherein the keys are used to
accomplish substitution. The process is achieved by
performing an XOR operation as also addition and MOD
functions between the key and raw pixels. This operation is
iterated twice over image row values to ensure strong
encryption. The output cipher is sensitive to minor changes in
the value of the input pixels.
In [85], image scrambling-based encryption is presented.
Scrambling is carried out based on a random sequence, which
is generated using a formula of two inputs that act as a secret
key. Then, the pixels‘ locations are rearranged utilizing the
generated sequence. Permutation based encryption is suggested
in [86]. In this method, the Linear Feedback Shift Register
(LFSR) produces a PRN that is employed to generate two
shuffled images. Permutation of the image rows creates the
first one, while permutation of image columns forms the
second image. Both images are XORed in the final step to
create the encrypted image. The Chaotic Gravitational Search
(CGS) concept for encryption is introduced in [87]. In this
system, a binary gravitational search algorithm selects a
random PRNG, thus starting an encryption key among three
algorithms: Mersenne Twister (MT), SIMD-Oriented Fast
Mersenne Twister (SFMT), and Combined Multiple Recursive
(MRG) algorithms. Then, the Chaotic Gravitational Search
Algorithm (CGSA) produces the initial value for the selected
generator. The generated keystream is used to encrypt images
by applying permutation alone or in combination with the
substitution process.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
373 | P a g e
www.ijacsa.thesai.org
Authors in [88] suggest transposition and substitution
encryption-based techniques that utilize two pseudorandom
generators: an altered version of the Sophie Germain Prime
Generator (ASGPG) and Lehmer Random Number Generator
(LRNG), which generate a group of random numbers. The first
group is used to assign new values to image pixels by XORing
the random numbers with image pixels: a substitution
condition. The second group attains the transposition by
swapping the positions of pixels. In [89], improved Baker map
and LM are used to achieve image encryption. A two-
dimensional Baker chaotic map is employed to control the
chaotic LM parameters as well as the state variable. It is used
to increase the randomness and unpredictability levels. Two
random sequences are produced by the chaotic LM. The
sequences are then exploited to perform image encryption by
first shuffling the pixels, followed by substitution. Likewise,
in[90] chaos-based encryption is proposed, which follows the
permutation and substitution methodology to encrypt the
original image. Clifford‘s chaotic system and LM are iterated
for a particular value followed by a quantization process. Next,
positions are shuffled, and grayscale substitution is performed
as a final stage. Parameters of the Clifford system map or the
attractor, act as encryption keys.
In [91], the authors propose a Hyper-Chaos (HC) based
image encryption algorithm. They make use of a 5-D multi-
wing hyper-chaotic system to resist cryptanalysis. It generates
a chaotic sequence that is used first for pixel-level permutation
and then for bit-level permutation. Up to this point, confusion
is fulfilled. To achieve diffusion, the chaotic sequence
generated earlier is XORed with the confusion step output. It is
worth mentioning that the function‘s parameters are deduced
from the plain image properties. In [92] presents a unified
image encryption algorithm. This technique utilizes
Substitution-Box (S-Box) to realize the diffusion concept.
Hence, a keyed-piecewise linear chaotic map creates a key
stream sequence. This key is employed for diffusion operations
that are based on a specific formula between this key value and
image pixels. However, the intermediate encrypted image is
180° rotated and then scrambled before making the second
diffusion. In this technique, the decryption process is identical
to the encryption process. Table IV summarizes the related
references for the image encryption methods.
C. Region based Steganography
In the following paragraphs, some techniques will be
presented, wherein steganography is carried out in certain areas
of the cover image to primarily enhance security. Some
techniques exploit only image edges, while others use edges
and smooth areas. Image edges and boundaries are the areas in
which the intensity value changes sharply within a short
distance, while the smooth area is the area with low-intensity
variations. In [18], edge pixels are detected and selected for
embedding using the Canny edge detector. Huffman code is
applied to the secret bits in the primary step for data
compression. Then, the edge pixels are randomized, and the
number of pixels intended for embedding is determined by
coherent bit length L. Next, 2k correction is applied to achieve
better imperceptibility. Edge detection incorporated with
morphological dilation as part of the steganographic technique
is suggested in [20]. In this method, the secret message is
embedded in the image sharp regions, which are detected by
the Canny edge operator and optimized by the morphological
dilation operator. The Canny operator is applied to a modified
version of the original image channels obtained by adding the 4
MSBs of RGB channels. Subsequently, a 3x3 dilation operator
is utilized to identify reference pixels. The bits are inserted
within the remaining LSB bits using the hybrid XOR
technique, such that least possible alterations of edge pixels are
guaranteed. This meets the high security demands.
TABLE IV. SUMMARY OF ENCRYPTIONS TECHNIQUES OF IMAGE ENCRYPTION TECHNIQUES
Ref
Encryption Concept
[78]
Confusion based encryption using Arnold Cat map and HM algorithm
[79]
Encryption is achieved by rotation, modified zigzag transformation, and enhanced LM output with XOR operation. It is a confusion and diffusion
encryption.
[80]
Bit-level image Permutation using a chaotic skew tent function
[81]
Permutation using a chaotic LM in addition to DNA encoding operations
[82]
Arnold transformation encryption of DNA
[83]
Permutation-substitution encryption utilizing intertwining logistic, enhanced LM, and cosine-based transformation
[84]
Substitution by exploiting 1D-dimensional cosine fractional (1-DCF) chaotic map
[85]
Scrambling based encryption using random sequence based on user formula
[86]
Permutation based encryption using linear feedback shift registers along with XOR operation
[87]
Permutation only or with substitution by exploiting a Binary gravitational search algorithm.
Mersenne twister, SIMD-oriented Fast Mersenne twister, and combined multiple recursive.
[88]
Transposition and substitution encryption based on Sophie Germain prime generator and Lehmer random number generator
[90]
Transposition and substitution encryption using an improved baker map and LM
[91]
Permutation - substitution encryption utilizing Clifford chaotic system and LM
[92]
Permutation - substitution encryption using a 5-D multi-wing hyper-chaotic system
[92]
Substitution - permutation encryption by employing a keyed-piecewise linear chaotic map

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
374 | P a g e
www.ijacsa.thesai.org
In [93], the secret data is hidden in the edges of the image
using the LSB technique. The Canny edge detection algorithm
is employed to identify image edges. The secret message is
converted into a binary sequence and encrypted using the OTP
encryption method. The encryption process is achieved by
modulus addition of the secret bits with the corresponding OTP
bits. Laplacian of Gaussian (LoG) detector is used in [94]
alongside a chaotic function. The secret information is first
encrypted by XORing it with a PRN generated by a chaotic
LM. Then, the edges of the green and blue planes are identified
using the LoG edge operator. Finally, 2-LSB of the green and
blue edge planes are used for the embedding process. In [19],
the author suggests using a hybrid edge detector and encryption
to secure the steganographic technique. The Vernam cipher is
applied to the secret message using a pseudo-random generated
key that is generated using a nonlinear feedback shift register,
Geffe Generator. This method uses a hybrid edge detector
which combines the Sobel operator with Kirch operators. The
LSB technique is used to hide secret message bits; here, three
bits are embedded in the edge pixels while two bits are
embedded in non-edge pixels.
In [95], an Adaptive Multi Bit-Planes image steganography
using Block Data-Hiding (MPBDH) is proposed. First, the
secret message is encrypted using the AES algorithm. Then,
complex regions are chosen for embedding, based on a
complexity threshold estimation and the number of bit planes.
If the whole message cannot be embedded within the selected
pixels, then parameter readjustment is used to compulsorily
ensure that all bits are embedded. LSB and Hamming code-
based algorithm is suggested in [96]. The cover image is
divided into blocks, and then edge detection is performed after
zeroing all LSBs of the red channel pixels. The embedding is
done based on the pixel‘s location. If the pixel is a non-edge
pixel, standard LSB embedding is used for the RGB channels.
Otherwise, LSB is performed for the three channels and the
other two LSBs of the RG channels are used to embed the
Hamming code. In [97], the complexity is based on the pixel
variation among the central pixel and all neighboring pixels.
The cover image is divided into small overlapping blocks
(3x3). Then, a multidirectional high pass filer bank is used to
find eight residual responses, which are then utilized to
calculate the corresponding complexity using a proposed
Complex Block Prior (CBP) criterion. The blocks are also
sorted from high to low complexity, and the blocks with the
same complexity level are grouped together. Finally, a single
bit or multiple bits are embedded adaptively within the central
pixel.
The Block-Wise Edge Adaptive Steganography Scheme
(BEASS) method is suggested in [98]. It utilizes the edges
obtained by using a fuzzy edge detector as well as the
surrounding pixels to embed secret message bits. The cover
image is divided into 64x64 blocks, and their corresponding
standard deviation is calculated to estimate their local
complexity. The blocks are sorted according to their standard
deviation, and then the blocks are further divided into 3x3
blocks. Three message bits are hidden in edge pixels using the
minimal mean squared error (MSE) that helps determine the
embedding capacity of neighboring non-edge pixels within the
block. Table V summarizes the related references about the
region-based image steganography methods.
TABLE V. SUMMARY OF REGION-BASED STEGANOGRAPHY TECHNIQUES
Ref
Features and Pros
Cons
PSNR (dB)
Evaluations Metrics
[18]
 Secret message compressed using Huffman
coding and randomly embedded in edge pixels
(Canny edges)
 Coherent bit length L determines the number of
embedding pixels
 2k correction maintains visual quality
 Huffman Table must be exchanged
 Steganalysis evaluation is missing
 Steps add complexity
 Limited capacity
512×512 gray-scale
Payloads =35852:108074
bits
Avg. PSNR= 61.9654
PSNR, Universal
Image Quality Index
(Q)
[19]
 Vernam cipher encryption of the secret message
with a pseudo-random key to enhance security
 Efficient hybrid edge detector of Sobel and Kirch
operators to improve capacity
 LSB adaptive embedding in Green & Blue
channels.
 7 bits of 2 block of 3x3 pixels utilized as
embedding indicator
 Steganalysis evaluation is missing
 Encoding complexity
90x90 secret image
512x512 cover image
Avg bpp=1.9605
Avg PSNR= 41.533
MSE, PSNR, bpp
[20]
 Canny edge detector with dilation operator to
involve edge pixels and their neighbours to
improve capacity
 Embedding using an improved XOR technique to
improve security
 Robust against (RS) steganalysis
 Relatively Low embedding rate
 Encoding complexity.
 Only RS steganalysis tested

512x512 Greyscale and
color images
Avg bpp=1.25
Avg PSNR = 44
MSE, PSNR, SSIM,
FSIM
[93]
 To add security, the secret message is OTP
encrypted
 OTP key is obtained using random function
 LSB based in Canny edge pixels.
 Handling of message key and host image
edges
 Steganalysis evaluation is missing
 Low capacity
512x512 greyscale images
Payload: Up to 1KB
PSNR for 1KB =69.1106
CC, PSNR, MSE
[94]
 LoG edge pixels of green and blue channels is
used for embedding process
 Efficient low complexity XOR encryption of the
message with chaotic logistic map sequence
 Parameters exchange overhead
 Steganalysis evaluation is missing
 Relatively Low capacity as 2 channels are
used
512x512 RGB image
bpp=1.72
PSNR= 46.1733
PSNR, NCC, Entropy,
Number-of-changes-
per-rate, and Unified-
average changed-
intensity.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
375 | P a g e
www.ijacsa.thesai.org
[95]
 The secret message is AES encrypted and
adaptively embedded in noisy regions
 Muti bit planes is utilized with block data-hiding
 Enhanced image visual quality
 Robust against visual attack and ensemble
classifier
 RSA key exchange overhead
 Other well-known steganalysis are not
tested
 Complexity is based on the bit planes not
value of pixels
 Ensemble classifier test for low bpp
512x512 gray images
bpp=0.4, PSNR= 56.64,
wPSNR= 71.96
bpp=1.50,
PSNR=48.08
wPSNR=63.20
PSNR, wPSNR, SSIM,
Visual attack,
Ensemble classifier
[96]
 Adaptive embedding based on LSB and
Hamming code.
 Non-edge pixel use LSB embedding
 Canny edge pixel to embed secret bits and
Hamming code
 Imperceptibility improved with adaptive
embedding
 Steganalysis evaluation is missing
 Few tests
 Capacity affected by Hamming code
 Cleared red LSBs affects the visual
quality
RGB 512x512, 512x384
from USC-SIPI, UCID-
Image Database
Payload: 9424:30224bits
PSNR: 63.397 :68.429
PSNR, MSE
[97]
 A content-adaptive steganography method based
on identifying the local texture complexity.
 Complexity based on pixel variation with respect
to all neighbours using multi-directional High
Pass Filter bank
 The maximum of the pixel differences in a pixel
block determines number of embedding bits
 Embedding using LSBMR or multibit XOR
 Too local complexity estimation since
small blocks size are utilized
 Other steganalysis tests not tested
 Relatively complex
 Parameter handling
512x512 greyscale SIPI,
BOWS2 and BOSSbase
dataset
Max capacity= 249,729:
549,051 (0.95: 2.09 bpp)
PSNR= 44.16:56.23
wPSNR=63.88: 75.87
Capacity, PSNR,
WPSNR and SSIM
SPAM
STEGANALYSIS
[98]
 Block-wise Edge Adaptive Steganography
Scheme (BEASS)
 Dynamic local complexity measure of Standard
Deviation is utilized (image subblocks containing
edges are selected)
 high payload with minimal distortion embedding
utilizing the minimal Mean Square Error
 Robust against major attacks
 Standard deviation as a complex measure
is not accurate for big blocks as it does
not take into account the spatial
arrangement of pixels
 wPSNR measure is more accurate in such
cases
 Low embedding capacity
 High complexity
512x512 greyscale images
from BOSSbase database
bpp= 0.3 : 1
PSNR (0.3 bpp) =70.54:
74.66
PSNR (∼1 bpp) = 61.28:
65.78
PSNR, KL-
Divergence, SSIM,
Average Difference,
NAE, Execution time,
Kurtosis, Skewness,
Histogram analysis,
RS attacks, and feature
based universal
steganalyzer
D. Bit-Plane System
The following LSB steganography techniques use the
concept of virtual bit-plane. Higher plane systems are exploited
instead of using an 8-bit plane to hide secret data. In these
systems, the Zeckendorf criterion needs to be satisfied in order
to embed the secret information that can be extracted later.
―Every positive integer can be represented uniquely as the sum
of one or more non-consecutive distinct Fibonacci numbers‖
[99]. Hence, the number representation is considered valid if
no consecutive ones appear in the sequence.
In the scheme presented in [99], secret data is embedded in
different bit-planes rather than using the regular binary bit
planes. It is based on the Lucas Number system that uses 11
numbers for representations. Hence, the Lucas sequence is
utilized for image bit plane representations, which uses 11 bits
instead of 8 bits for representing the pixel‘s intensity.
Embedding is achieved in the second bit-plane, which yields
deterioration by ±1 in the stego image. Blue and green
channels of the RGB color image are used for data embedding,
while red is used as an indicator. As mentioned, embedding
should comply with the extended Zeckendorf theorem for
handling redundant representation. In [100], the method
represents the cover image using the Fibonacci sequence. The
cover image in this situation is represented in 12-bit planes.
The secret message to be embedded is encrypted using a
symmetric cipher with the help of a chaotic LM to generate the
encryption key. Then, the generated key is XORed with secret
bits and embedded in the second LSB.
Authors in [101] suggest improving the Fibonacci data
hiding technique by utilizing Catalan numbers. A certain set of
Fibonacci numbers and a certain set of Catalan numbers are
combined to create a new number set that complies with
Zeckendorf‘s theorem. As a result, 15 virtual bit-planes are
obtained, which is three more than the number of bit-planes
produced by the Fibonacci method. Authors in [102] propose
two embedding techniques by utilizing two different number
systems. The first embedding scheme is based on a prime
decomposition of pixel value into 15 virtual bit-planes. In
contrast, the second is based on natural number decomposition,
which yields 23 virtual bit-planes. In these techniques, the
secret message can be embedded in higher bit-planes without
introducing noticeable distortion. In [103], a new method based
on a specific representation is used to decompose pixel
intensity values into 16 virtual bit-planes. This scheme
necessitates that the sum of all bit-planes must be less than the
highest pixel intensity value, which in this case is 28-1. For the
embedding process, a cover pixel is randomly selected using
PRNG, then decomposed, and a particular virtual-bit plane is
chosen. To represent numbers that have multiple
representations, the one with high lexicographically
representation is selected. In addition, if a pixel value post
embedding cannot be represented in the system, it is excluded,
as the embedded information cannot be extracted. The author
in [104] proposes a steganographic technique based on a new
pixel value decomposition. This scheme uses a set of
decomposition weights that decompose the cover image into 10
bit-planes. The first five LSBs‘ weights increase slightly while
growing exponentially for the rest. The secret message bits are
embedded in one or more LSB positions. Embedding is only
done for modified pixels with a valid representation in the
system. Moreover, in several cases, numbers have more than
one representation in the system. Hence, the one with the
lowest lexicographical representation is a valid number. Table
VI summarizes the references related to bit-plane based
methods.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
376 | P a g e
www.ijacsa.thesai.org
TABLE VI. SUMMARY OF BIT-PLANE BASED STEGANOGRAPHY TECHNIQUES
Ref
Features and Pros
Cons
PSNR (dB)
Evaluations Metrics
[99]
 Using Lucas sequence to represent pixel‘s
intensity
 Yields deterioration by ±1 in the stego
image.
 Reduces visual distortion effects.
 Robust against geometrical, statistical and
structural attacks
 Low capacity than standard LSB
steganography
 Produce lower quality stego image
 Image structure not considered
 More complex than binary
representation
512x512, 394x600, 768x512,
200x200 RGB USC-SIPI &
Kodak
bpp=10%:100%
PSNR=57.27:47:34
MSE, PSNR, SSIM
histogram differences,
image, Chi-square
attack, structural attacks
and RS attack
[100]
 Cover image represented using Fibonacci
sequence
 XOR encryption of LM sequence with
secret message
 Embedding in the 2nd bit plane
 Simple and effective encryption using the
XOR operation
 Some pixels escaped
 Image structure not considered
 Steganalysis attacks missing
 Computation overhead
 Parameters exchange overhead
512x512 greyscale images
from CVG-UGR Database
bpp=0.3, PSNR = 54.21
bpp=0.9, PSNR = 41.04
Avg PSNR :49.97
ER, MSE, PSNR
Robustness tests
Bit error rate BER
rate-distortion curve
[101]
 Number set composed of union of a
certain set of Fibonacci numbers and a
certain set of Catalan numbers
 15 virtual bit-planes obtained
 Robust against statistical and geometrical
attacks
 Some pixels escaped (affects capacity)
 Image structure not considered
 Steganalysis attacks missing
 Computation overhead
Greyscale images
Payloads: 1000:2500 bits
PSNR= 67.03: 71.45
MSE, PSNR
[102]
 Cover image can be represented using
prime sequence (15 bits), or using natural
number sequence (23 bits)
 Secret message can be embedded in higher
bit-planes without introducing distortion.
 Image structure not considered
 Steganalysis attacks missing
 Computation overhead
 Low visual quality for higher bit planes
Greyscale images
(Natural, prime)
PSNR(bit plane): 0th = 48, 48,
1st = 43, 43, 7th = 30, 24
Worst case Mean Square
Error (WMS)
PSNR, Histogram
[103]
 Pixels randomly selected and decomposed
into 16 bit-planes.
 A particular bit plane is chosen
 Produces less distortion
 Some pixels cannot be used for
embedding
 Capacity is affected by unusable pixels
 Image structure not considered
 Steganalysis attacks missing
512 x 512
PSNR:
1st LSB= 52, 2nd LSB= 50,
7th LSB= 37, 8th LSB= 34
Payload capacity, PSNR
[104]
 Cover image is decomposed into 10 virtual
bit-planes
 The first 5 LSB weight increases slightly
 Embedding in one or more LSBs positions
(valid pixels)
 Robust against statistical
 Not every pixel is valid after
embedding
 Capacity is affected by unusable pixels
 Image structure not considered
 Steganalysis attacks missing
 Computation overhead
512x512 jpg images
Embedding using: 0th,1st,
2nd,3rd,4th bit
PSNR: 39.41: 49.02
MSE, PSNR, SSIM
E. Pixel Indicator Techniques
In indicator-based data embedding schemes, the channels
are divided into indicator and data channels. The indicator
channel determines the data channel for data hiding, based on
certain sequences for better security. In [16], the two least
significant bits of one channel are used as an indicator of the
presence of secret data in the other two data channels. The
indicator channel is chosen based a sequence created from R,
G, and B, that is, RGB, RBG, GBR, GRB, BRG, and BGR.
The length of the secret message is used as an indicator
selection criterion to enhance security. In addition, hiding data
in channels considers the pixel‘s intensity.
In [1], the 7th bit of a pixel and the 7th bit of the pixel value
+1 are concluded. A comparison between those values and two
bits from the secret message is conducted to seek a matching.
In the case of a mismatch, the difference is calculated and then
added to the pixel value. At this point, the embedding process
is completed for this pixel. Eventually, the embedding process
alters the pixel value by +2 or -2. The aim of designing this
algorithm is to implement robust and secure steganography. In
[25], a variant version of PIT is presented. This technique uses
indicator channel criteria to determine the data channel order,
based on the combinations of the MSB in the RGB channel.
The first step is to obtain the message length, which is then
stored in the first 8 bytes of the first row. Then, embedding is
started from the second row, and the indicator channel
selection criteria is utilized to determine the order of the data
channel. The selection criteria are based on the message length
and the type of the parity bit used to create a shuffled order of
RGB channels. Also, the number of bits to be embedded in
each channel is specified.
The study [105] suggests a PIT based technique to achieve
high capacity. Here, the binary secret message is divided into
four parts and the cover image. The two LSB of the first eight
bytes of the red channel of the image are used to store the
message length. The order of the cover image parts is stored in
the two least significant bits of the four first pixels in the blue
channel. Next, each pixel is evaluated for the ability to embed
using the red channel‘s three MSBs. The three bits represent
the RGB ability: the presence of zero means no embedding has
taken place in that channel. Then, the number of zeros in the
four MSBs of the candidate channel is counted to determine
the number of secret bits to be hidden. The method proposed in
[106] utilizes the red channels as an indicator to hide the secret
message in the fifth and sixth bits of either the green or the
blue channels of the cover image. The count of ones in the red
channel determines which channel is currently in use. If the
count is even, the green channel is used to embed the secret
data; otherwise, the blue is used. A similar technique is found
Table VIII. Summary of PIT based steganography techniques
Ref
Features and Pros
Cons
Evaluations
Metrics
[1]

The embedding is based on the indicator bit, the 7th
bit of a pixel value p and p+1

Robust and secure steganography

Simple implementation
The embedding process alters the pixel value by +2
or -2.

Lack of steganalysis.

Subject to statistical steganalysis

All image regions are considered affects
security and imperceptibility

Low capacity
256x256 from USC-
SIPI-ID
dataset.
Payload: 2: 10 KB
PSNR: 55.39: 48.39
bpp=0.031: 0.16
PSNR, MSE,
Histogram
[16]

The indicator channel selection is secret message
length dependant to enhance security

Produces low visual distortion

Hiding process takes into account the pixel‘s
intensity
Standard PIT algorithm uses 2-bits LSB‘s

It uses a fixed number of bits per channel
that could cause noticeable distortion

Uses fixed embedding sequence

Image structure not considered

Steganalysis evaluation is missing
512×384 BMP image
2: 8KB, 256x256
avg PSNR= 43.65:
52.81
PSNR, Histogram,
Mean, standard
deviations
[17]

Indicator channel selected based on the message
length

Data channel is selected according to 3LSBs of
indicator channel

Improved security due to using indicator with
multi-mode indicators with adaptive channel
embedding

Image structure not considered

Steganalysis attacks missing

Limited capacity due pixels escaping
RGB images
Payloads: 1Kb, 2Kb,
4Kb
Avg. PSNR:64.45,
62.45, 60.45
PSNR, MSE
[97]

It uses 3 MSB of red channel as an indicator

Number of zero bits in MSB determines storage
capacity

To enhance security, input image divided to 4
regions and each part of secret message will store
in this region.
Some pixels are used to store control bits

Image structure not considered

Steganalysis evaluation is missing

Subject to statistical and geometrical
attacks
RGB with different
sizes
2000 character
2.144122 bpp (avg)
avg PSNR 60:57
Histogram, Mean,
standard deviations,
PSNR
[98]

Uses 5th & 6th LSBs to enhance security since
attacker focus on LSB bits for secret data
extraction

Number of ones in the indicator channel
determines the data channel

2 bits of the other channels are used as data bits.

Secret message bits are XORed with predefined
secret key to increase security

All image contents are considered

Steganalysis evaluation is missing

Subject to structural and geometrical
attacks

Low payload capacity
RGB with different
sizes
PSNR (512X512) =
54.65
Max payload =2 bpp
PSNR
Histogram
Mean values

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
377 | P a g e
www.ijacsa.thesai.org
in [107] in which the green channel is used as an indicator. If
the count of the number of ones in the green channel is even,
the red channel is used. Otherwise, the blue one is used.
Utilizing two bits for embedding process of each pixel leads to
reasonable payload capacity but enhances its security. Table
VII summarizes the references related to PIT based techniques.
F. Steganographic Techniques Based on Combination of
Different Concepts
In this section, several techniques are seen to apply
different concepts to achieve secure steganography. In the
scheme proposed in [108], Huffman coding is applied to the
secret message to reduce its size. Then, based on the location
(x, y) of the current pixel, either a bitwise XNOR operation or
Fibonacci algorithm is applied to embed the secret message. If
x is less than y, the XNOR operation is employed to embed the
secret bits in the green or blue channels, and the red channel is
used as an indicator. If y is greater than x, the Fibonacci
algorithm is applied to embed secret bits in the second LSB of
the Fibonacci virtual green bit plane. Otherwise, the pixel is
skipped. In [109] the author suggests a steganography
technique based on control bit and chaotic bit stream. A chaotic
LM is used to generate a chaotic bit stream. Then, this stream
is XORed with the LSB of the cover image pixels to create a
control bit matching with the corresponding secret message bit.
Therefore, the embedding process may change the LSB value
or be kept intact based on specific criteria. A technique based
on Adaptive Directional Pixel Value Differencing (ADPVD) is
suggested in [110]. The aim is to enhance embedding capacity
and security. The method starts by evaluating the PVD
embedding capacity by traversing three directions: the
horizontal, vertical, and diagonal directional edges. That is
accomplished after partitioning the original image into two-
pixel non-overlapping blocks. The highest capacity rate is
calculated for each color channel, and then the appropriate
direction is selected.
In [111], the authors suggest a steganography technique
based on variant expansion and modulus function. The purpose
is to improve embedding capacity as well as the security of the
stego image. The cover image is divided into non-overlapping
blocks of two pixels. The proposed method considers multiple
directions for each color channel and adaptively selects the
appropriate embedding direction that achieves the highest
embedding capacity. Instead of only considering positive
differences, this method selects positive and negative
difference values to hide secret data. In [112], the authors
present a steganographic method for an RGB image based on
the PVD technique. As in most PVD based techniques, the
cover image is partitioned into sequential non-overlapping
blocks of two pixels. Then, two-color channels combinations
are created: red and green color components for the first
combination, and green and blue color components for the
second. Next, the secret data is embedded in each block using
the PVD technique and then examined and readjusted to obtain
the modified three-color combination. This operation utilizes a
particular threshold to control the embedding capacity of each
block in order to minimize distortion.
TABLE VII. SUMMARY OF PIT BASED STEGANOGRAPHY TECHNIQUES
Ref
Features and Pros
Cons
Evaluations
Metrics
[1]
 The embedding is based on the indicator bit, the 7th
bit of a pixel value p and p+1
 Robust and secure steganography
 Simple implementation
 The embedding process alters the pixel value by +2
or -2.
 Lack of steganalysis.
 Subject to statistical steganalysis
 All image regions are considered affects
security and imperceptibility
 Low capacity
256x256 from USC-
SIPI-ID
dataset.
Payload: 2: 10 KB
PSNR: 55.39: 48.39
bpp=0.031: 0.16
PSNR, MSE,
Histogram
[16]
 The indicator channel selection is secret message
length dependant to enhance security
 Produces low visual distortion
 Hiding process takes into account the pixel‘s
intensity
 Standard PIT algorithm uses 2-bits LSB‘s
 It uses a fixed number of bits per channel
that could cause noticeable distortion
 Uses fixed embedding sequence
 Image structure not considered
 Steganalysis evaluation is missing
512×384 BMP image
2: 8KB, 256x256
avg PSNR= 43.65:
52.81
PSNR, Histogram,
Mean, standard
deviations
[25]
 Indicator channel selected based on the message
length
 Data channel is selected according to 3LSBs of
indicator channel
 Improved security due to using indicator with multi-
mode indicators with adaptive channel embedding
 Image structure not considered
 Steganalysis attacks missing
 Limited capacity due pixels escaping

RGB images
Payloads: 1Kb, 2Kb,
4Kb
Avg. PSNR:64.45,
62.45, 60.45
PSNR, MSE
[105]
 It uses 3 MSB of red channel as an indicator
 Number of zero bits in MSB determines storage
capacity
 To enhance security, input image divided to 4
regions and each part of secret message will store in
this region.
 Some pixels are used to store control bits
 Image structure not considered
 Steganalysis evaluation is missing
 Subject to statistical and geometrical
attacks
RGB with different
sizes
2000 character
2.144122 bpp (avg)
avg PSNR 60:57
Histogram, Mean,
standard deviations,
PSNR
[106]
 Uses 5th & 6th LSBs to enhance security since
attacker focus on LSB bits for secret data extraction
 Number of ones in the indicator channel determines
the data channel
 2 bits of the other channels are used as data bits.
 Secret message bits are XORed with predefined
secret key to increase security
 All image contents are considered
 Steganalysis evaluation is missing
 Subject to structural and geometrical
attacks
 Low payload capacity
RGB with different
sizes
PSNR (512X512) =
54.65
Max payload =2 bpp
PSNR
Histogram
Mean values

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
378 | P a g e
www.ijacsa.thesai.org
A steganographic method based on an LM function
combined with LSB and PVD to improve security is suggested
in [21]. A random key is generated using a chaotic LM that is
employed to pick two pixels randomly at a time. In addition,
the key value is utilized along with the MOD function to
operate either with 3-LSB substitution or PVD, to embed the
secret information. In [113], the author claims a novel
steganographic approach known as Clustering Modification
Directions (CMDs). In this scheme, instead of focusing only on
the embedding location clusters of the texture area, clustering
of the direction modifications is considered as well. The
purpose is to obscure statistical features to resist steganalysis.
±1 LSB embedding approach is considered in this situation.
The cover image is segmented into several sub-images as well
as the secret message. After embedding the first message
portion, the costs of the adjacent segments are updated to
cluster the embedding direction. Each time embedding occurs,
the costs are updated. A variable-length group of bits
substitution based scheme is presented in [114]. In this scheme,
embedding for a Group of Bits‘ Substitution (GBS) is done by
replacing a group of bits in a pixel with another group of bits of
the same message length. The scheme is designed to work in
two variations: 1-bit GBS and 2-bit GBS, which hide 1-bit and
2-bits, respectively. The choice is based on certain predefined
conditions. Image imperceptibility and security are improved
since most of the pixel values remain unchanged.
The authors in [115] suggest LSB based steganography
with Optical Character Recognition (OCR). The concept is
based on using a secret message in the form of characters
within an image. Character-level features are extracted from
the secret image and then embedded into a cover image using
the standard LSB. The OCR model has been developed and
trained to extract character features. The security perspective of
this technique is that, even if the attacker extracts the
embedded bits, he still needs to know the OCR model in order
to recover the original message. In [116], the authors propose a
new method by combining the Right-Most Digit Replacement
(RMDR) with an Adaptive Least Significant Bit (ALSB). The
cover image is divided into lower texture and higher texture
regions. Accordingly, either RMDR or ALSB is chosen to
embed the secret message based on RMD rather than bits.
RMDR is employed to embed secret bits in the lower texture
regions, whereas ALSB is used in high texture regions.
In [117], the RGB cover image is cropped into a predefined
number of crops with certain secret coordinates. This number
is used to divide the secret text message accordingly. Each part
of the message is then embedded using standard LSB into a
certain cropped part using a secret sequence. Embedding is
achieved using the 3-3-2 sequence. Finally, stego crops are
assembled to create the stego image. The coordinates of
cropped parts are considered as the key and agreed upon
between the two parties. The approach presented in [118] uses
two images: a reference image and a cover image. The
reference image is divided into N blocks, wherein every block
is assigned a unique code. The secret message is also divided
into N-bits blocks that are encoded using the block codes
obtained earlier. If there is no match between the secret block
and the block codes, some LSBs of the reference image need to
be altered. Information such as the starting block and traversing
direction is incorporated into a secret key, which is then
encrypted using the RSA algorithm. Finally, the encoded bit
sequence is embedded into the cover image using any LSB
technique.
Utilizing bit plane indexes, authors in [119] suggest a
secure steganography technique. In this scheme, the secret
message is embedded in multiple image bit planes to enhance
security without sacrificing capacity payload. It is based on
manipulating bit planes indexes. Only the two LSB bits are
employed for this purpose. The cover image is initially
preprocessed such that the first two bits are not be equal. If the
first LSB bit equals the first secret bit, the index is recorded. If
they mismatch, the second LSB plane is recorded. In the next
turn, the recorded index is in reverse order, for example, if it
previously recorded zero, it is one the next time and alteration
to LSB is done accordingly. Hence, the final index stream
fluctuates between zero and one. In [120], a different
perspective is developed to achieve image steganography.
Instead of modifying separate image pixels, which causes
random noise in the image, this technique changes the image‘s
color palette. All pixels of the same color are transformed into
the same color. Therefore, this method achieves a higher user
perception. Utilizing quad-trees, authors in [29] present a
steganographic method in luminance (L* channel) and
chrominance (a* and b* channels) (L*a*b) color space. This
approach utilizes a quad-tree segmentation process to partition
the spatial domain of the cover image into high correlation and
low correlation adaptive size blocks. Embedding is done in the
high frequency regions of the DCT of the highly correlated
cover image blocks. To improve stego quality, L*a*b color
space is utilized. A high quality stego image is guaranteed
along with better security, since the embedding takes place
only in the high frequency regions that produce minimum
image degradation. The performance of this method is affected
by the correlation of the image, wherein a highly correlated
image is preferred. Table VIII summarizes the related
references.
TABLE VIII. SUMMARY OF STEGANOGRAPHIC TECHNIQUES BASED ON DIFFERENT CONCEPTS
Ref
Features and Pros
Cons
PSNR (dB)
Evaluations metrics
[21]
 High capacity and Security by combining
LSB&PVD with LM to randomly select two
consecutive pixels
 Using either LSB or PVD based on mod
function and LM
 3 bits embedded in case of LSB
 Lack of steganalysis.
 Low visual quality.
 All image regions are considered
affects security and imperceptibility
512×512 greyscale
Payloads:
bpp=2.26: 2.37
Avg PSNR: 38.7925
PSNR, Histogram
analysis
[29]
 Quad-tree utilized to obtain High & low
correlations adaptive-size blocks
 Embedding only in high frequency regions
 The performance is affected by the
correlation of the image (highly
correlated image is preferred
512 x 512 color images with
variety of correlations
Capacity 76%-90%
SSIM, Combined
Capacity Quality
Effective-ness

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
379 | P a g e
www.ijacsa.thesai.org
DCT of chrominance channels (a*b*)
 Hence high-quality and better security is
guaranteed
 Robust against low-density attacks
 Moderate PSNR
 Payload capacity not mentioned in the
steganalysis
Avg PSNR: 37.317
(CCQE).
Attacks: filtering,
geometric, and
compression
attacks.
[108]
 Hybrid bit planes (Fibonacci) with XNOR
operation
 Huffman coding to compress the secret
message.
 Effective simple encryption offers high
security
 High imperceptibility
 Image structure not considered
 Steganalysis attacks missing
 Computation overhead
 Huffman table exchange overhead
RGB 512x512 images
Payloads: 8MB & 16KB
PSNR: 65.153:74.192
PSNR, MSE,
Embedding
capacity, Histogram
[109]
 XORing LSB with chaotic bitstream to produce
control bit
 Embedding based on comparison of LSB with
Control bit
 Simple implementation
 Secure against brute force attack
 System parameters exchange overhead
 Image structure not considered
 Steganalysis attacks missing
 Subject to statistical steganalysis
300x300, 512x512,
1024x1024 greyscale images
Payloads: 512: 8192 bytes
PSNR= 51.96: 64.82
Correlation
coefficient, entropy,
PSNR, and
Image Fidelity
[110]
 Blocks of two non-overlapping pixels
 Adaptively selects the direction for each color
channel
 Simple implementation.
 Enhancement of embedding capacity and
security
 Image structure not considered
 Steganalysis attacks missing
 Moderate visual quality
 Three direction calculation - overhead
512x512
bpp =1.65 (greyscale)
PSNR=46.71
bpp = 1.63 (RGB)
PSNR=51.59
Capacity, PSNR,
histogram analysis
[111]
 Based on variant expansion and modulus
function
 Adaptive embedding direction with positive
and negative differences are considered
 Simple implementation with enhanced capacity
and imperceptibility
 Overhead of parameters exchange
 All image regions are considered
 Steganalysis attacks missing
 Moderate capacity
512x512 RGB from SIPI
Highest PSNR= 52.216
Vertical & bpp=1.573
Highest capacity: Diagonal
bpp=1.609
MSE, NPCR,
embedding capacity,
NPCR, UACI, and
pixel difference
histogram analysis
[112]
 RGB-PVD based scheme.
 PVD of the two overlapping channel
combination of (R, G) and (G, B) with
readjustment of the RGB components
 Capacity is improved due to overlapping
blocks
 Low visual quality (Low PSNR)
 No steganalysis evaluation
 Sequential embedding and all image
regions are considered affects security
and imperceptibility
512x512 RGB images:
bpp=2.53
PSNR= 32.79
PSNR, MSE,
Payload capacity
[113]
 Modifications are considered along with
clustering the directions (+ or -) of embedding
modification by updating the cost
 Robust against high-dimensional features and
ensemble classifiers
 Can be used together with schemes with
additive distortion functions, such as HILL, S-
UNIWARD, WOW
 High complexity
 Works with sub images
 The costs of pixels within each sub-
image are dynamically adjusted
512x512 gray-scale images
from BOSSBase image
database.
Testing
Classification error
Steganalytic
performance
(maxSRMd2)
Steganalytic
performance
(tSRM)
[114]
 A variable-length group of bits substitution-
based scheme with two variations (1-bit & 2-
bits)
 Image imperceptibility and security are
improved since the majority of pixel values
remain unchanged
 Lack of steganalysis.
 Subject to statistical steganalysis
 All image regions are considered
affects imperceptibility
 Moderate capacity
512×512 RGB images
1 bit: Payloads= 1 bpp
PSNR=51.64
2 bits: Payloads=2 bpp
PSNR= 49.762
PSNR, hiding
capacity, image
quality index, and
pixel difference
histograms
[115]
 Enhanced security since character features of
text in secret images are used as secret message
 Need of a trained OCR model.
 standard LSB.
 Efficient in training time and accuracy (SMO
classifier)
 All image regions are considered
(affects imperceptibility)
 Computationally expensive
 Overhead of classifier training and
testing
 Preprocessing steps overhead
 Character-Feature table handling
RGB Steganalysis Dataset.
Payloads: 128×128 grey
images.
PSNR:
51.107 (1 bpp)
43.094 (2 bpp)
36.444 (3 bpp)
PSNR, MSE, SSIM
[116]
 Combining RMDR with ALSB
 The RMDR & ALSB offer high embedding
capacity and maintain a good imperceptibility
and Security
 Texture complexity utilized to use either
RMDR or ALSB
 Robust against statistical steganalysis.
 Low block size used to determine
texture level.
 SPAM + ensemble classifier can
successfully steganalyze for a higher
embedding rate
 Not tested against structural detectors
 Moderate PSNR
512x512, 256x256,
1024x1024 from UCID,
USC-SIPI
bpp = 3.052
PSNR= 39.00
PSNR, Q, RS-
analysis, pixel
difference histogram
analysis, and SPAM
features under
ensemble classifier
steganalysis

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
380 | P a g e
www.ijacsa.thesai.org
[117]
 Uses 3-3-2 sequence and cropping cover image
into k parts and the secret message is divided
into k parts
 Embedding using 3-3-2 sequence
 Security is related to number of parts,
coordinates, and sequence pattern
 Image structure not considered
 Steganalysis evaluation is missing
 Subject to statistical and structural
steganalysis
 Fixed embedding sequence
512x512 RGB images
PSNR: 62.5332
PSNR, MSE
[118]
 Message encoding using two images divided
into k blocks.
 Secret message encoded using reference image
and then embedded in the LSB
 Secret key encrypted using RSA involves:
starting block, traversing direction, etc
 Image structure not considered
 Steganalysis evaluation is missing
 Subject to statistical and structural
steganalysis
 Two images needed
 Key exchange overhead
256x256
Payload= 2000 bits
Avg. PSNR = 71.48
Payload= 16000 bits
Avg. PSNR = 62.36
PSNR, MSE,
Histogram
[119]
 Manipulates bit-planes indexes to enhance
security.
 2 LSB should be in the form 01,10. The cover
image is pre-processed accordingly and the
vector of indices to be sent
 Robust against PoV, WS steganalysis
 Handling of the vector of indices
 All image contents are considered
 Modern steganalysis can attack
 Not robust against MLSB-WS
steganalyser when bpp=1
512 x 512 Greyscale
Payload:
20%: 100%
PSNR: ~47 for bpp=1
PSNR, PoV, WS
steganalysis,
MLSB-WS
steganalysis
[120]
 Changes the color palette
 All pixels of the same color are changed to the
same color
 Achieves a higher user perception.
 Allows resistance to analysis of adjacent pixel
colors
 Its capacity is very dependent on a
color palette
 Not resistant to color palette analysis
and standard palette images
 Need to test over modern Steganalysis
512 × 512
bpp = 0.35: 1.37
PSNR: 52.74:58.10
PSNR, SSIM, EC
VI. OBSERVATIONS, DISCUSSION, AND RECOMMENDATIONS
A. Observations
 Chaotic based randomness: even though chaotic based
LSB steganography achieves higher security, the
payload capacity attained is low, and there exists low
robustness against statistical and geometric attacks.
 Secret message encryption-based approach: this concept
provides higher security, but the complexity is high,
especially while using substitution-permutation
encryption.
 Image encryption: using standard encryption
techniques, multilevel encryption techniques, and
chaotic based techniques provides good security.
However, the overall system overhead is high.
 Virtual multi-bit plane-based steganography: this
paradigm achieves better payload capacity and higher
security as the possibility of randomness is greater.
Nonetheless, the secret data can deteriorate if there is a
slight stego image change by attackers. It is particularly
vulnerable to non-statistical steganalysis (geometrical
attacks) such as rotation, scaling, and cropping.
 Region based steganography: these techniques achieve
good robustness and security, but the embedding
capacity in general is low.
B. Discussion
Hiding secret information inside a cover image without
introducing suspicious artefacts is the main objective of image
steganography. In addition, high security, good
imperceptibility, and high embedding rate are desirable and
challengeable goals. Researchers have been working on
enhancing the performance of steganographic algorithms in
terms of achieving high security, high imperceptibility, and
high payload capacity. Yet, the optimum goal has not been
reached since the challenging criteria oppositely affect each
other. When the main consideration is security, the technique
should hide the presence of embedded data inside the cover
image from the attacker‘s attention. In addition, it should
obscurely hide the data so that attackers cannot identify the
original secret message even if they detect its presence. Several
concepts have recently been utilized to achieve high security in
image steganography.
For securing steganographic techniques, the concept
employed most widely is encryption, which has been used to
add a layer of security. Encryption can be attained utilizing
standard encryption techniques such as AES, 3DES, and RSA
along with secret keys to enhance the overall system security.
On the other hand, user-defined encryption algorithms are also
utilized via the concepts of permutation only, substitution only,
or both. The level of security can also be boosted much by
embedding the secret data in non-sequential order, that is, by
using random sequences to scatter the secret bits all over the
cover image. Existing chaotic functions are famous for
producing random numbers that can be exploited to create
random sequences. Further, some researchers rely on varied
concepts to generate such randomness. Pixel channels indicator
is another paradigm that has been used to indicate the presence
or absence of secret information and identifies the color
channel being used.
The visual aspects of an image are also utilized to achieve a
level of security, since the image is composed of smooth
regions and non-smooth regions also known as low frequency
and high frequency regions. Embedding data in a smooth area
can raise distortion levels; this breaches the confidentiality of
secret data. On the other hand, exploiting the non-smooth
regions and edges as embedding locations does not leave
evidence of the existence of secret data. Employment of
number systems to create virtual bit planes is another means of
hiding the secret information without creating noticeable
distortion as well as of hiding pixels‘ relations. Such attributes
overcome the security limitations of the standard spatial
domain techniques. Frequency domain steganography is
another approach that has been followed to guarantee the

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
381 | P a g e
www.ijacsa.thesai.org
security of steganography by choosing appropriate locations to
embed secret data. This technique avoids manipulating pixels
directly and instead uses transform procedures, thereby leading
to good imperceptibility. However, the embedding capacity is
limited and the computational cost is higher.
C. Recommendations
The recommendations of this study are as follows:
 Combining edge-based steganography with
randomness-based concepts to achieve higher security
approaches that resist statistical steganalysis;
 Utilizing the existing encryption methods to add an
extra layer of security;
 Applying the adaptiveness concept by combining
multiple hiding techniques based on some image
attributes or user-defined criteria (techniques such as
quad tree search are useful to segregate the cover image
into various segments with different attributes);
 Instead of utilizing the entire image to embed the secret
data, hiding data in particular regions known as the
Region of Interest (ROI), which resists statistical
attacks by breaking the statistics relations of adjacent
pixels;
 Employing the optimization concept to enhance the
security of chaotic based steganography (machine
learning techniques are an appropriate method to
achieve such a goal);
 Drawing more attention to images in the YCBCR color
system, since it has received less attention in this
context; and
 Considering 3D for embedding secret data, as very few
attempts have been made in this domain.
VII. CONCLUSION
Image steganography is used to hide secret information
inside a cover image. It is frequently used to guarantee
confidentiality while sending information over an untrusted
network. A comprehensive review of image steganography in
the spatial domain was carried out utilizing recent research
mainly through IEEE Explore, ScienceDirect, Springer Link,
and other databases. Most methods utilize LSB based
steganography in the spatial domain and its variants due to its
simplicity and effectiveness. In addition, PVD, EMD, PIT have
been employed as well. In this work, a review of image
steganography in the spatial domain in general and, more
precisely, of the security aspect, led to its classification into
multiple categories. These categories are as follows:
randomization based, encryption based, randomization and
encryption based steganographic techniques, image encryption,
region-based steganography, multiple bit-planes based, pixel
indicator techniques, and other steganographic techniques
based on combinations of different concepts. At the end of this
review, some discussion of the gaps and the future scope of the
above-mentioned concepts has been included. In addition,
future recommendations for new and current researchers
interested in this field are provided.
ACKNOWLEDGMENT
This work was supported by Universiti Kebangsaan
Malaysia under research grant GP-2021-K011439.
REFERENCES
[1] K. Joshi, S. Gill, and R. Yadav, ―A New Method of Image
Steganography Using 7th Bit of a Pixel as Indicator by Introducing the
Successive Temporary Pixel in the Gray Scale Image,‖ Journal
ofComputer Networks and Communications, vol. 2018, p. 10, 2018, doi:
https://doi.org/10.1155/2018/9475142.
[2] O. H. Alhabeeb, F. Fauzi, and R. Sulaiman, ―A Review of Modern
DNA-based Steganography Approaches,‖ IJACSA , vol. 12, no. 10,
2021, doi: 10.14569/IJACSA.2021.0121021.
[3] M. A. Majeed, R. Sulaiman, Z. Shukur, and M. K. Hasan, ―A review on
text steganography techniques,‖ Mathematics, vol. 9, no. 21. MDPI,
Nov. 01, 2021. doi: 10.3390/math9212829.
[4] M. A. Majeed and R. Sulaiman, ―An improved LSB image
steganography technique using bit-inverse in 24 bit colour image,‖ J
Theor Appl Inf Technol, vol. 80, no. 2, 2015.
[5] A. H. Ali, L. E. George, A. A. Zaidan, and M. R. Mokhtar, ―High
capacity, transparent and secure audio steganography model based on
fractal coding and chaotic map in temporal domain,‖ Multimed Tools
Appl, vol. 77, no. 23, pp. 31487–31516, Dec. 2018, doi:
10.1007/s11042-018-6213-0.
[6] S. Kamil, M. Ayob, Siti Norul Huda Sheikh Abdullah, and M. Zulkifli
Ahmad, ―Lightweight and Optimized Multi-Layer Data Hiding using
Video Steganography,‖ (IJACSA), vol. 9, no. 12, 2018, doi:
10.14569/IJACSA.2018.091237.
[7] N. Subramanian, O. Elharrouss, S. Al-Maadeed, and A. Bouridane,
―Image Steganography : A Review of the Recent Advances,‖ IEEE
ACCESS, vol. 9, pp. 23409–23423, 2021, doi:
10.1109/ACCESS.2021.3053998.
[8] P. C. Mandal, I. Mukherjee, G. Paul, and B. N. Chatterji, ―Digital image
steganography: A literature survey,‖ Inf Sci (N Y), vol. 609, pp. 1451–
1488, Sep. 2022, doi: 10.1016/j.ins.2022.07.120.
[9] A. Rashid and M. K. Rahim, ―Critical analysis of stegauography ‗An art
of hidden writing,‘‖ International Journal of Security and its
Applications, vol. 10, no. 3, pp. 259–282, 2016, doi:
10.14257/ijsia.2016.10.3.24.
[10] R. Article, A. K. Sahu, and M. Sahu, ―Digital image steganography and
steganalysis : A journey of the past three decades,‖ Open Computer
Science, vol. 10, no. 1, pp. 296–342, 2020, doi:
https://doi.org/10.1515/comp-2020-0136.
[11] I. J. Kadhim, P. Premaratne, P. J. Vial, and B. Halloran,
―Comprehensive survey of image steganography: Techniques,
Evaluations, and trends in future research,‖ Neurocomputing, vol.
Volume 335, pp. 299-326, 2019, doi: 10.1016/j.neucom.2018.06.075.
[12] K. Muhammad, M. Sajjad, I. Mehmood, S. Rho, and S. W. Baik, ―A
Novel Magic LSB Substitution Method ( M-LSB-SM ) using Multi-
Level Encryption and Achromatic Component of an Image,‖ Multimed
Tools Appl 75, 14867–14893 (2016). https://doi.org/10.1007/s11042-
015-2671-9.
[13] M. C. Alipour, B. D. Gerardo, and R. P. MEDINA, ―LSB Substitution
Image Steganography Based on Randomized Pixel Selection and One-
Time Pad Encryption,‖ BDSIC 2020: 2020 2nd International
Conference on Big-data Service and Intelligent ComputationDecember
2020 Pages 1–6https://doi.org/10.1145/3440054.3440055, pp. 1–6.
[14] C. Chan and L. M. Cheng, ―Hiding data in images by simple LSB
substitution,‖ Pattern Recognit, vol. 37, no. 3, pp. 469–474, 2004, doi:
10.1016/j.patcog.2003.08.007.
[15] J. Mielikainen, ―LSB matching revisited,‖ IEEE Signal Process Lett,
vol. 13, no. 5, pp. 285–287, May 2006, doi: 10.1109/LSP.2006.870357.
[16] F. Huang and J. Huang, ―Edge Adaptive Image Steganography Based on
LSB Matching Revisited,‖ IEEE TRANSACTIONS ON INFORMATION
FORENSICS AND SECURITY, vol. 5, no. 2, 2010, doi:
10.1109/TIFS.2010.2041812.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
382 | P a g e
www.ijacsa.thesai.org
[17] G. Liu, Z. Zhang, Y. Dai, and S. Lian, ―Improved LSB-matching
steganography for preserving second-order statistics,‖ J Multimed, vol.
5, no. 5, pp. 458–463, Oct. 2010, doi: 10.4304/jmm.5.5.458-463.
[18] S. Sun, ―A Novel edge based image steganography with 2k correction
and Huffman encoding,‖ Inf Process Lett, no. September, 2015, doi:
10.1016/j.ipl.2015.09.016.
[19] Z. F. Yaseen, ―Image Steganography Based on Hybrid Edge Detector to
Hide Encrypted Image using Vernam Algorithm,‖ in 2nd Scientific
Conference of Computer Sciences (SCCS), University of Technology -
Iraq Image, IEEE, 2019, pp. 75–80.
[20] K. Gaurav and U. Ghanekar, ―Journal of Information Security and
Applications Image steganography based on Canny edge detection ,
dilation operator and hybrid coding,‖ Journal of Information Security
and Applications, vol. 41, pp. 41–51, 2018, doi:
10.1016/j.jisa.2018.05.001.
[21] S. Prasad and A. K. Pal, ―Logistic Map-Based Image Steganography
Scheme Using Combined LSB and PVD for Security Enhancement,‖
Advances in Intelligent Systems and Computing, vol. 814, pp. 203–214,
2019, doi: 10.1007/978-981-13-1501-5.
[22] M. Tools, K. Bailey, and K. Curran, ―An Evaluation of Image Based
Steganography Methods,‖ International Journal of Digital Evidence,
vol. 2, no. 2, pp. 55–88, 2003, doi: 10.1007/s11042-006-0008-4.
[23] K. Muhammad, J. Ahmad, N. U. Rehman, Z. Jan, and R. J. Qureshi, ―A
Secure Cyclic Steganographic Technique for Color Images using
Randomization,‖ Technical Journal, University of Engineering and
Technology Taxila, vol. 19, no. III, pp. 57–64, 2014.
[24] A. A. Gutub, ―Pixel Indicator Technique for RGB Image
Steganography,‖ JOURNAL OF EMERGING TECHNOLOGIES IN
WEB INTELLIGENCE, vol. 2, no. 1, pp. 56–64, 2010, doi:
10.4304/jetwi.2.1.56-64.
[25] J. Pandey, K. Joshi, M. Jangra, and M. Sain, ―Pixel Indicator
Steganography Technique with Enhanced Capacity for RGB Images,‖ in
2019 International Conference on Intelligent Computing and Control
Systems (ICCS), IEEE, 2019, pp. 738–743.
[26] X. Zhang and S. Wang, ―Efficient steganographic embedding by
exploiting modification direction,‖ IEEE Communications Letters, vol.
10, no. 11, pp. 781–783, 2006, doi: 10.1109/LCOMM.2006.060863.
[27] Z. S. Younus and M. K. Hussain, ―Image steganography using
exploiting modification direction for compressed encrypted data,‖
Journal of King Saud University – Computer and Information Sciences,
vol. 34, no. 6, pp. 2951–2963, Jun. 2019, doi:
https://doi.org/10.1016/j.jksuci.2019.04.008.
[28] S. Venkatraman, A. Abraham, and M. Paprzycki, ―Significance of
steganography on data security,‖ International Conference on
Information Technology: Coding Computing, ITCC, vol. 2, pp. 347–351,
2004, doi: 10.1109/ITCC.2004.1286660.
[29] M. Baziyad, T. Rabie, and I. Kamel, ―L * a * b * color space high
capacity steganography utilizing quad-trees,‖ Multimedia Tools and
Applications (2020), pp. 25089–25113, 2020.
[30] G. G. Rajput and Ramesh Chavan, ―A Novel Approach for Image
Steganography Based On Random LSB Insertion in Color Images,‖ in
Proceedings of the International Conference on Intelligent Computing
Systems, 2017, pp. 265–273. doi:
https://dx.doi.org/10.2139/ssrn.3131654.
[31] M. G. Gouthamanaath, ―Hiding Three Binary Images in a Grayscale
Image with Pixel Matching Steganography and Randomization
technique,‖ in Proceeding of 2018 IEEE International Conference on
Current Trends toward Converging Technologies, Coimbatore, India
Hiding, IEEE, 2018, pp. 1–8.
[32] M. G. Gouthamanaath and A. Kangaiammal, ―Hiding binary image in a
grayscale image using Pixel Matching and Randomization Technique,‖
in Proc. of the Fourth Intl. Conf. Advances in Computing,
Communication and Information Technology- CCIT 2016, 2016, pp. 74–
78. doi: 10.15224/978-1-63248-092-7-30.
[33] U. A. E. Ali, ―A LSB Based Image Steganography Using Random Pixel
and Bit Selection for High Payload,‖ no. August, pp. 24–31, 2021, doi:
10.5815/ijmsc.2021.03.03.
[34] J. L. Pichardo Méndez, L. Palacios Luengas, R. F. Martínez González,
O. Jiménez Ramírez, and R. Vázquez Medina, ―LSB Pseudorandom
Algorithm for Image Steganography Using Skew Tent Map,‖ Arab J Sci
Eng, vol. 45, no. 4, pp. 3055–3074, 2020, doi: 10.1007/s13369-019-
04272-0.
[35] C. Pak, J. Kim, K. An, C. Kim, K. Kim, and C. Pak, ―A novel color
image LSB steganography using improved 1D chaotic map,‖ Multimed
Tools Appl, vol. 79, no. 1–2, pp. 1409–1425, 2020, doi:
https://doi.org/10.1007/s11042-019-08103-0.
[36] Z. Rim, A. Afef, E. Ridha, and Z. Mourad, ―Beta Chaotic Map Based
Image Steganography,‖ in 12th International Conference on
Computational Intelligence in Security for Information Systems (CISIS
2019) and 10th International Conference on EUropean Transnational
Education (ICEUTE 2019), Springer, Cham, 2019, pp. 97–104. doi:
10.1007/978-3-030-20005-3.
[37] A. V. Gahan and G. D. Devanagavi, ―A Secure Steganography Model
Using Random-Bit Select Algorithm,‖ in 2020 Third International
Conference on Advances in Electronics, Computers and
Communications (ICAECC), Bengaluru, Dec. 2020. doi:
10.1109/ICAECC50550.2020.9339474.
[38] M. A. F. Al-Husainy and D. M. Uliyan, ―A SECRET-KEY IMAGE
STEGANOGRAPHY TECHNIQUE USING RANDOM CHAIN
CODES,‖ International Journal of Technology, vol. 10, no. 4, pp. 731–
740, 2019, doi: https://dx.doi.org/10.14716/ijtech.v10i4.653.
[39] S. Dagar, ―Highly Randomized Image Steganography using Secret
Keys,‖ in IEEE International Conference on Recent Advances and
Innovations in Engineering, IEEE, 2014, pp. 1–5. doi: doi:
10.1109/ICRAIE.2014.6909116.
[40] S. A. Nie, G. Sulong, R. Ali, and A. Abel, ―The use of least significant
bit ( LSB ) and knight tour algorithm for image steganography of cover
image,‖ vol. 9, no. 6, pp. 5218–5226, 2019, doi:
10.11591/ijece.v9i6.pp5218-5226.
[41] E. Alrashed and S. S. Alroomi, ―Hungarian-Puzzled Text with Dynamic
Quadratic Embedding Steganography,‖ vol. 7, no. 2, pp. 799–809, 2017,
doi: 10.11591/ijece.v7i2.pp799-809.
[42] M. Saidi, H. Hermassi, R. Rhouma, and S. Belghith, ―LSB-Hamming
based Chaotic Steganography,‖ in 12th International Conference for
Internet Technology and Secured Transactions (ICITST), Cambridge,
UK: IEEE, 2017, pp. 29–34. doi: 10.23919/ICITST.2017.8356340.
[43] L. Laimeche, A. Meraoumia, and H. Bendjenna, ―Enhancing LSB
Embedding Schemes Using Chaotic Maps Systems,‖ Neural Comput
Appl, vol. 32, pp. 16605–16623, 2020, doi:
https://doi.org/10.1007/s00521-019-04523-z.
[44] H. Elkamchouchi, Wessam M. Salama, and Yasmine Abouelseoud,
―Data Hiding in a Digital Cover Image using Chaotic Maps and LSB
Technique,‖ in 12th International Conference on Computer Engineering
and Systems (ICCES), Cairo, Egypt, 2017. doi:
10.1109/ICCES.2017.8275302.
[45] S. Mukherjee and G. Sanyal, ―A chaos based image steganographic
system,‖ Multimed Tools Appl, vol. 77, no. 21, pp. 27851–27876, 2018.
[46] S. S. Shankar and A. Rengarajan, ―Puzzle based Highly Secure
Steganography,‖ in International Conference on Algorithms,
Methodology, Models and Applications in Emerging Technologies
(ICAMMAET), 2017. doi: 10.1109/ICAMMAET.2017.8186742.
[47] B. Mondal, ―A Secure Steganographic Scheme Based on Chaotic Map
and DNA Computing,‖ In: Sharma, D.K., Balas, V.E., Son, L.H.,
Sharma, R., Cengiz, K. (eds) Micro-Electronics and Telecommunication
Engineering. Lecture Notes in Networks and Systems, vol. 106, pp. 545–
554, 2020, doi: https://doi.org/10.1007/978-981-15-2329-8_55.
[48] S. SOLAK and U. ALTINIŞIK, ―A New Approach for Steganography:
Bit Shifting Operation of Encrypted Data in LSB (SED-LSB),‖ Journal
of information technology, vol. 12, no. 1, pp. 75–82, 2019, doi:
10.17671/gazibtd.435437.
[49] A. Setyono, D. R. Ignatius, and M. Setiadi, ―Securing and Hiding Secret
Message in Image using XOR Transposition Encryption and LSB
Method,‖ in Journal of Physics: Conference Series, Volume 1196,
International Conference on Information System, Computer Science and
Engineering, 2019. doi: 10.1088/1742-6596/1196/1/012039.
[50] R. Dumre and A. Dave, ―Exploring LSB Steganography Possibilities in
RGB Images,‖ in 12th International Conference on Computing

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
383 | P a g e
www.ijacsa.thesai.org
Communication and Networking Technologies, ICCCNT 2021,
Kharagpur: IEEE, 2021. doi: 10.1109/ICCCNT51525.2021.
[51] S. Chauhan, Jyotsna, J. Kumar, and A. Doegar, ―Multiple layer Text
security using Variable block size Cryptography and Image
Steganography,‖ in 3rd International Conference on Computational
Intelligence & Communication Technology (CICT), IEEE, 2017. doi:
https://doi.org/10.1109/CIACT.2017.7977303.
[52] M. M. Abdel-aziz, K. M. Hosny, and N. A. Lashin, ―Improved data
hiding method for securing color images,‖ Multimed Tools Appl, vol. 80,
pp. 12641–12670, 2021, doi: https://doi.org/10.1007/s11042-020-10217-
9.
[53] G. Maji, S. Mandal, and S. Sen, ―Cover independent image
steganography in spatial domain using higher order pixel bits,‖
Multimed Tools Appl, vol. 80, pp. 15977–16006, 2021, doi:
https://doi.org/10.1007/s11042-020-10298-6.
[54] S. Zhang, L. Yang, X. Xu, and T. Gao, ―Secure Steganography in JPEG
Images Based on Histogram Modification and Hyper Chaotic System,‖
vol. 10, no. 1, pp. 40–53, 2018, doi: 10.4018/IJDCF.2018010104.
[55] S. Rahman et al., ―A Novel Approach of Image Steganography for
Secure Communication Based on LSB Substitution Technique,‖
Computers, Materials & Continua, vol. 64, no. 1, pp. 31–61, 2020, doi:
10.32604/cmc.2020.09186.
[56] F. B. Calanda, A. M. Sison, M. R. D. Molato, and R. P. Medina, ―A
Modified Least Significant Bit Randomized Embedding Method based
on Image Partitioning and Columnar Transposition with Encryption,‖ in
ICCBD 2019: Proceedings of the 2nd International Conference on
Computing and Big Data, ACM, 2019, pp. 68–72. doi:
https://doi.org/10.1145/3366650.3366662.
[57] O. S. Sitompul, Z. Situmorang, F. R. Naibaho, and E. B. Nababan,
―STEGANOGRAPHY WITH HIGHLY RANDOM LINEAR
CONGRUENTIAL GENERATOR FOR SECURITY
ENHANCEMENT,‖ 2018 Third International Conference on
Informatics and Computing (ICIC), pp. 1–6, 2018, doi:
10.1109/IAC.2018.8780445.
[58] B. Srinivasan, S. Arunkumar, and K. Rajesh, ―A Novel Approach for
Color Image , Steganography Using NUBASI and Randomized , Secret
Sharing Algorithm,‖ Indian J Sci Technol, vol. 8, no. April, pp. 228–
235, 2015, doi: 10.17485/ijst/2015/v8iS7/.
[59] Z. S. Younus and M. K. Hussain, ―Image steganography using
exploiting modification direction for compressed encrypted data,‖
Journal of King Saud University – Computer and Information Sciences,
2019, doi: https://doi.org/10.1016/j.jksuci.2019.04.008.
[60] D. Ghosh, A. K. Chattopadhyay, and A. Nag, ―A Novel Approach of
Image Steganography with Encoding,‖ In: Chakraborty, M.,
Chakrabarti, S., Balas, V., Mandal, J. (eds) Proceedings of International
Ethical Hacking Conference 2018. Advances in Intelligent Systems and
Computing, vol. 811, pp. 115–124, 2019, doi: 10.1007/978-981-13-
1544-2.
[61] S. Elshare and N. N. El-emam, ―Modified Multi-Level Steganography to
Enhance Data Security,‖ International Journal of Communication
Networks and Information Security (IJCNIS), vol. 10, no. 3, pp. 509–
525, 2018.
[62] H. R. Kareem, H. H. Madhi, and K. A. Mutlaq, ―Hiding encrypted text
in image steganography,‖ Periodicals of Engineering and Natural
Sciences, vol. 8, no. 2, pp. 703–707, 2020.
[63] M. N. Abdulwahed, ―An effective and secure digital image
steganography scheme using two random function and chaotic map,‖ J
Theor Appl Inf Technol, vol. 98, no. 1, pp. 78–91, 2020.
[64] K. Kordov and S. Zhelezov, ―Steganography in color images with
random order of pixel selection and encrypted text message embedding,‖
PeerJ Comput. Sci., pp. 1–21, 2021, doi: 10.7717/peerj-cs.380.
[65] P. Yadav and M. Dutta, ―3-Level Security Based Spread Spectrum
Image Steganography with Enhanced Peak Signal to Noise Ratio,‖ in
2017 Fourth International Conference on Image Information Processing
(ICIIP), IEEE, 2017, pp. 122–126. doi: 10.1109/ICIIP.2017.8313696.
[66] M. C. E. M. Kasapbaşi, ―A New Chaotic Image Steganography
Technique Based on Huffman Compression of Turkish Texts and Fractal
Encryption With Post-Quantum Security,‖ in IEEE Access, vol. 7, pp.
148495–148510, 2019, doi: 10.1109/ACCESS.2019.2946807.
[67] R. F. Martínez-González, J. A. Díaz-Méndez, L. Palacios-Luengas, J.
López-Hernándezd, and R. Vázquez-Medina, ―A steganographic method
using Bernoulli‘ s chaotic maps,‖ Computers and Electrical
Engineering, vol. 54, no. C, pp. 435–449, 2016, doi:
10.1016/j.compeleceng.2015.12.005.
[68] M. Kasapbas and W. Elmasry, ―New LSB-based colour image
steganography method to enhance the efficiency in payload capacity ,
security and integrity check,‖ Sādhanā, vol. 43, no. 68, 2018, doi:
https://doi.org/10.1007/s12046-018-0848-4.
[69] H. Alhelow, ―Highly Secure Steganography-Based System with Three
Layers of Protection,‖ EasyChair, 2021.
[70] U. I. Assad and G. M. Bhat, ―Hiding in encrypted images : a three tier
security data hiding technique,‖ Multidimens Syst Signal Process, 2015,
doi: 10.1007/s11045-015-0358-z.
[71] E. J. Kusuma, C. A. Sari, E. H. Rachmawanto, and D. R. I. M. Setiadi,
―A Combination of Inverted LSB , RSA , and Arnold Transformation to
get Secure and Imperceptible Image Steganography,‖ Journal of ICT
Research and Applications, vol. 12, no. 2, pp. 103–122, 2018, doi:
10.5614/itbj.ict.res.appl.2018.12.2.1.
[72] A. Hussain and P. Bora, ―A Highly Secure Digital Image Steganography
Technique Using Chaotic Logistic Map and Support Image,‖ in
Proceedings of 2018 IEEE International Conference on Information
Communication and Signal Processing (ICSP 2018) A, IEEE, 2018, pp.
69–73.
[73] J. S. Neenu and E. B. Varghese, ―A novel approach for SCC algorithm
using pattern based image steganography,‖ Proceedings of the
International Conference on Inventive Computation Technologies,
ICICT 2016, vol. 2016, pp. 1–6, 2016, doi:
10.1109/INVENTIVE.2016.7830241.
[74] P. Das, S. C. Kushwaha, and M. Chakraborty, ―Data Hiding Using
Randomization and Multiple Encrypted Secret Images,‖ in 2015
International Conference on Communications and Signal Processing
(ICCSP), 2015, IEEE, 2015, pp. 298–302. doi:
10.1109/ICCSP.2015.7322892.
[75] K. Tiwari and S. J. Gangurde, ―LSB Steganography Using Pixel Locator
Sequence with AES,‖ 2021 Second International Conference on Secure
Cyber Computing and Communication (ICSCCC), vol. 255, pp. 302–
307, 2021, doi: 10.1109/ICSCCC51823.2021.9478162.
[76] R. Shanthakumari, S. Varadhaganapathy, S. Vinothkumar, and B.
Bharaneeshwar, ―Data hiding in Image steganography using Range
Technique for secure communication,‖ in 2021 International
Conference on Advances in Electrical, Computing, Communication and
Sustainable Technologies (ICAECT), 2021. doi:
https://doi.org/10.1109/ICAECT49130.2021.9392480.
[77] T. K. Hue, N. T. Linh, M. Nguyen-duc, and T. M. Hoang, ―Data Hiding
in Bit-plane Medical Image Using Chaos-based Steganography,‖ in 2021
International Conference on Multimedia Analysis and Pattern
Recognition (MAPR), IEEE, 2021. doi:
https://doi.org/10.1109/MAPR53640.2021.9585243.
[78] R. K. Sinha, ―Chaotic Image Encryption Scheme Based on Modified
Arnold Cat Map and Henon Map,‖ 2018 International Conference on
Current Trends towards Converging Technologies (ICCTCT), pp. 1–5,
2018.
[79] P. Ramasamy, V. Ranganathan, S. Kadry, and R. Damaševiˇ, ―An Image
Encryption Scheme Based on Block Scrambling , Modified Zigzag
Transformation and Key Generation Using Enhanced Logistic—Tent
Map,‖ Entropy, vol. 21, no. 7, p. 656, 2019, doi: 10.3390/e21070656.
[80] L. Teng, X. Wang, and J. Meng, ―A chaotic color image encryption
using integrated bit-level permutation,‖ Multimed Tools Appl, vol. 77,
no. 6, pp. 6883–6896, 2017, doi: 10.1007/s11042-017-4605-1.
[81] B. Mondal and T. Mandal, ―A light weight secure image encryption
scheme based on chaos & DNA computing,‖ Journal of King Saud
University - Computer and Information Sciences, vol. 29, no. 4, pp. 499–
504, 2017, doi: 10.1016/j.jksuci.2016.02.003.
[82] Y. Yang, B. Guan, J. Li, D. Li, Y. Zhou, and W. Shi, ―Image
compression-encryption scheme based on fractional order hyper- chaotic
systems combined with 2D compressed sensing and DNA encoding,‖
Opt Laser Technol, vol. 119, no. November 2018, p. 105661, 2019, doi:
10.1016/j.optlastec.2019.105661.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
384 | P a g e
www.ijacsa.thesai.org
[83] B. kumar Nancharla and M. Dua, ―An Image Encryption using
Intertwining Logistic map and Enhanced Logistic Map,‖ in Fifth
International Conference on Communication and Electronics Systems
(ICCES 2020), IEEE, 2020, pp. 1309–1314.
[84] M. Z. Talhaoui, X. Wang, and A. Talhaoui, ―A new one-dimensional
chaotic map and its application in a novel permutation-less image
encryption scheme,‖ Vis Comput, 2020, doi: 10.1007/s00371-020-
01936-z.
[85] K.S.K.S.Sarma and B.Lavanya, ―Digital Image Scrambling based on
Sequence Generation,‖ in 2017 International Conference on Circuit
,Power and Computing Technologies (ICCPCT), 2017. doi:
10.1109/ICCPCT.2017.8074317.
[86] S. Saha, R. K. Karsh, and M. Amrohi, ―Encryption and Decryption of
Images using Secure Linear Feedback Shift Registers,‖ 2018
International Conference on Communication and Signal Processing
(ICCSP), pp. 295–298, 2018.
[87] B. O. Al-roithy and A. A. Gutub, ―Trustworthy image security via
involving binary and chaotic gravitational searching within PRNG
selections,‖ IJCSNS International Journal of Computer Science and
Network Security, vol. 20, no. 12, pp. 167–176, 2020, doi:
https://doi.org/10.22937/IJCSNS.2020.20.12.18.
[88] A. Ramesh and A. Jain, ―Hybrid Image Encryption using Pseudo
Random Number Generators , and Transposition and Substitution
Techniques,‖ in International Conference on Trends in Automation,
Communications and Computing Technology (I-TACT-15), IEEE, 2015.
doi: https://doi.org/10.1109/ITACT.2015.7492652.
[89] Y. Luo, J. Yu, W. Lai, and L. Liu, ―A novel chaotic image encryption
algorithm based on improved baker map and logistic map,‖ Multimed
Tools Appl, vol. 78, no. 15, Aug. 2019, doi: 10.1007/s11042-019-7453-
3.
[90] B. Fathi-vajargah, M. Kanafchian, and V. Alexandrov, ―Image
Encryption Based on Permutation and Substitution Using Clifford
Chaotic System and Logistic Map,‖ J Comput (Taipei), vol. 13, no. 3,
pp. 309–326, 2017, doi: 10.17706/jcp.13.3.309-326.
[91] Y. Li, C. Wang, and H. Chen, ―A hyper-chaos-based image encryption
algorithm using pixel-level permutation and bit-level permutation,‖ Opt
Lasers Eng, vol. 90, no. August 2016, pp. 238–246, 2017, doi:
10.1016/j.optlaseng.2016.10.020.
[92] Y. Zhang, ―The unified image encryption algorithm based on chaos and
cubic S-Box,‖ Inf Sci (N Y), vol. 450, pp. 361–377, 2018, doi:
10.1016/j.ins.2018.03.055.
[93] C. Irawan, D. R. I. M. Setiadi, C. A. Sari, and E. H. Rachmawanto,
―Hiding and securing message on edge areas of image using LSB
steganography and OTP encryption,‖ Proceedings - 2017 1st
International Conference on Informatics and Computational Sciences,
ICICoS 2017, vol. 2018-Janua, no. February 2018, pp. 1–6, 2017, doi:
10.1109/ICICOS.2017.8276328.
[94] A. Jan, S. A. Parah, and B. A. Malik, ―A Novel Laplacian of Gaussian (
LoG ) and Chaotic Encryption Based Image Steganography Technique,‖
in International Conference for Emerging Technology (INCET), IEEE,
2020. doi: https://doi.org/10.1109/INCET49848.2020.9154173.
[95] T. D. Nguyen, S. Arch-int, and N. Arch-int, ―An adaptive multi bit-plane
image steganography using block data-hiding,‖ Multimed Tools Appl,
vol. 75, pp. 8319–8345, 2016, doi: 10.1007/s11042-015-2752-9.
[96] Y. Wang, M. Tang, and Z. Wang, ―Optik High-capacity adaptive
steganography based on LSB and Hamming code,‖ Optik - International
Journal for Light and Electron Optics, vol. 213, no. March, p. 164685,
2020, doi: 10.1016/j.ijleo.2020.164685.
[97] A. Saeed et al., ―An accurate texture complexity estimation for quality-
enhanced and secure image steganography,‖ IEEE Access, vol. 8, pp.
21613–21630, 2020, doi: 10.1109/ACCESS.2020.2968217.
[98] D. Laishram and T. Tuithung, ―A novel minimal distortion-based edge
adaptive image steganography scheme using local complexity:
(BEASS),‖ Multimed Tools Appl, vol. 80, no. 1, pp. 831–854, Jan. 2021,
doi: 10.1007/s11042-020-09519-9.
[99] B. Datta, K. Dutta, and S. Roy, ―Data hiding in virtual bit-plane using
efficient Lucas number sequences,‖ Multimed Tools and applications,
vol. 79, pp. 22673–22703, 2020, doi: https://doi.org/10.1007/s11042-
020-08979-3.
[100] A. Rehman, T. Saba, T. Mahmood, Z. Mehmood, M. Shah, and A.
Anjum, ―Data hiding technique in steganography for information
security using number theory,‖ J Inf Sci, vol. 45, no. 6, pp. 767–778,
2018, doi: 10.1177/0165551518816303.
[101] N. Aroukatos, K. Manes, S. Zimeras, and F. Georgiakodis, ―Data Hiding
Techniques in Steganography using Fibonacci and Catalan numbers,‖ in
Ninth International Conference on Information Technology, IEEE, 2012.
doi: 10.1109/ITNG.2012.96.
[102] S. Dey, A. Abraham, B. Bandyopadhyay, and S. Sanyal, ―Data Hiding
Techniques Using Prime and Natural Numbers,‖ Journal of Digital
Information Management, vol. 6, no. 3, pp. 463–485, 2010, doi:
https://doi.org/10.48550/arXiv.1003.3672.
[103] A. A. Abdulla, H. Sellahewa, and S. A. Jassim, ―Steganography Based
on Pixel Intensity Value Decomposition,‖ in Proceedings Volume 9120,
Mobile Multimedia/Image Processing, Security, and Applications,
Baltimore, Maryland, United States, 2014. doi:
https://doi.org/10.1117/12.2050518.
[104] K. Biswas, ―A New Pixel Value Decomposition based Image
Steganography Method,‖ in 12th International Conference on
Computational Intelligence and Communication Networks, IEEE, 2020,
pp. 333–341. doi: 10.1109/CICN.2020.61.
[105] V. Rahmani and M. Mohammadpour, ―High hiding capacity
steganography method based on pixel indicator technique,‖ in 5th
Iranian Joint Congress on Fuzzy and Intelligent Systems - 16th
Conference on Fuzzy Systems and 14th Conference on Intelligent
Systems, CFIS 2017, Institute of Electrical and Electronics Engineers
Inc., Aug. 2017, pp. 144–149. doi: 10.1109/CFIS.2017.8003673.
[106] A. Sharma, M. Poriye, and V. Kumar, ―A Secure Steganography
Technique Using MSB,‖ International Journal of Emerging Research in
Management &Technology, vol. 6, no. 6, pp. 2278–9359, 2017, doi:
10.23956/ijermt.v6i6.270.
[107] S. Ahmed, R. Jaffari & Liaquat, and A. Thebo, ―Data Hiding Using
Green Channel as Pixel Value Indicator,‖ International Journal of
Image Processing (IJIP), vol. 12, no. 3, p. 90, 2018.
[108] A. A. Almayyahi, R. Sulaiman, F. Qamar, and A. E. Hamzah, ―High-
Security Image Steganography Technique using XNOR Operation and
Fibonacci Algorithm,‖ International Journal of Advanced Computer
Science and Applications, vol. 11, no. 10, pp. 511–522, 2020, doi:
10.14569/IJACSA.2020.0111064.
[109] H. Ogras, ―An Efficient Steganography Technique for Images using
Chaotic Bitstream,‖ I. J. Computer Network and Information Security,
vol. 11, no. 2, pp. 21–27, 2019, doi: 10.5815/ijcnis.2019.02.03.
[110] M. A. Hameed, S. Aly, and M. Hassaballah, ―An efficient data hiding
method based on adaptive directional pixel value differencing ( ADPVD
),‖ Multimed Tools Appl, vol. 77, pp. 14705–14723, 2018, doi:
10.1007/s11042-017-5056-4.
[111] M. ZULQARNAIN, M. G. GHOUSE, W. SHARIF, G. JILANIE, and A.
SHIFA, ―AN EFFICIENT METHOD OF DATA HIDING FOR
DIGITAL COLOUR IMAGES BASED ON VARIANT EXPANSION
AND MODULUS FUNCTION,‖ Journal of Engineering Science and
Technology, vol. 16, no. 5, pp. 4160–4180, 2021.
[112] S. Prasad and A. K. Pal, ―An RGB colour image steganography scheme
using overlapping block-based pixel value differencing,‖ R Soc Open
Sci, vol. 4, p. 161066, 2017, doi: http://dx.doi.org/10.1098/rsos.161066.
[113] B. Li, M. Wang, X. Li, S. Tan, and J. Huang, ―A strategy of clustering
modification directions in spatial image steganography,‖ IEEE
Transactions on Information Forensics and Security, vol. 10, no. 9, pp.
1905–1917, 2015, doi: 10.1109/TIFS.2015.2434600.
[114] G. Swain, ―Digital image steganography using variable length group of
bits substitution,‖ Procedia - Procedia Computer Science, vol. 85, pp.
31–38, 2016, doi: 10.1016/j.procs.2016.05.173.
[115] A. Chatterjee, S. K. Ghosal, and R. Sarkar, ―LSB based steganography
with OCR : an intelligent amalgamation,‖ Multimed Tools Appl, vol. 79,
pp. 11747–11765, 2020, doi: https://doi.org/10.1007/s11042-019-08472-
6.
[116] M. Hussain, A. W. A. Wahab, N. Javed, and K.-H. Jung, ―Hybrid Data
Hiding Scheme Using Right-Most Digit Replacement and Adaptive
Least Significant Bit for Digital Images,‖ Symmetry (Basel), vol. 8, no.
6, p. 41, 2016, doi: 10.3390/sym8060041.

(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 14, No. 6, 2023
385 | P a g e
www.ijacsa.thesai.org
[117] K. A. Al-afandy, E.-S. M. EL-Rabaie, O. S. Faragallah, A. Elmhalawy,
and Gh. M. El-Banby, ―High Security Data Hiding Using Image
Cropping and LSB Least Significant Bit Steganography,‖ in 2016 4th
IEEE International Colloquium on Information Science and Technology
(CiSt), Tangier, Morocco: IEEE, 2016, pp. 400–404. doi:
https://doi.org/10.1109/CIST.2016.7805079.
[118] G. Maji, S. Mandal, S. Sen, and N. C. Debnath, ―Dual Image based LSB
Steganography,‖ in 2018 2nd International Conference on Recent
Advances in Signal Processing, Telecommunications & Computing
(SigTelCom) Dual, IEEE, 2018, pp. 61–66. doi:
10.1109/SIGTELCOM.2018.8325806.
[119] A. A. Abdulla, S. A. Jassim, and H. Sellahewa, ―Secure Steganography
Technique Based on Bitplane Indexes,‖ in 2013 IEEE International
Symposium on Multimedia, IEEE, 2013, pp. 287–291. doi:
10.1109/ISM.2013.55.
[120] E. Margalikas and S. Ramanauskait, ―Image steganography based on
color palette transformation in color space,‖ EURASIP J Image Video
Process, pp. 1–13, 2019, doi: http://doi.org/10.1186/s13640-019-0484-x.