A Novel Image Steganography Method Based on Spatial Domain with

War Strategy Optimization and Reed Solomon Model

Hassan Jameel Azooz

1 a

, Khawla Ben Salah

2 b

, Monji Kherallah

3 c

and

Mohamed Saber Naceur

4 d

University of Almuthanna, Iraq

National Engineering School Sfax, Tunisia

Faculty of Sciences of Sfax, Tunisia

University of Carthage, Tunisia

Keywords:

War Strategy Optimization, Data Encryption, Data Security, Visual Similarity.

Abstract:

In this paper, we propose a novel approach to steganography using the War Search Optimization (WSO) al-

gorithm. Steganography is the practice of concealing messages within other data, such as images or audio

ﬁles. Our approach employs the WSO algorithm to optimize the parameters of a steganography algorithm,

aiming to maximize the perceptual similarity between the cover image and the stego image. We demonstrate

the effectiveness of our approach on a variety of cover images and secret messages and show that our method

produces stego images with high perceptual similarity to the cover images. Our results suggest that the WSO

algorithm is a promising tool for optimizing steganography algorithms. Also, this paper presents a new ap-

proach to steganography that utilizes the War Search Optimization (WSO) algorithm. Steganography involves

hiding messages within other data, such as images or audio ﬁles. Our method applies the WSO algorithm to

optimize the parameters of a steganography algorithm with the goal of maximizing the perceptual similarity

between the cover image and the stego image. We evaluate our approach on various cover images and secret

messages and demonstrate that our technique generates stego images with high perceptual similarity to the

cover images. The results indicate that the WSO algorithm is a valuable tool for optimizing steganography

algorithms.

1 INTRODUCTION

The application of evolving technology across a wide

range of scientiﬁc disciplines inevitably increases the

complexity of the challenges that must be addressed.

Due to the limitations of previous optimization meth-

ods, the metaheuristic optimization algorithm has

emerged as a viable alternative for solving difﬁcult

engineering problems. Therefore, modern optimiza-

tion algorithms provide some optimization because

of their advantages such as robustness, performance

reliability, simplicity, and ease of implementation.

Higher education institutions and workplaces today

rely heavily on written materials such as papers, ofﬁ-

cial letters, books, maps, etc., so the risk of forgery in-

https://orcid.org/0009-0004-4310-9310

https://orcid.org/0000-0002-4227-9623

https://orcid.org/0000-0002-4549-1005

https://orcid.org/0009-0001-4609-0086

creases due to the availability and simplicity of tech-

nology that may produce near-perfect copies of doc-

uments that contain sensitive information. To deal

with this problem, security in communication across

(the Internet, networks, etc.) for these documents has

become vital in order to prevent leakage of sensitive

information during transmission, and for this reason,

masking and encryption were used to secure the data.

Steganography is one of the areas of information se-

curity and is the art of concealing conﬁdential infor-

mation by embedding it in host media such as text,

image images, audio and video in order to protect it

from discovery or retrieval by unauthorized entities.

The image chosen to include the conﬁdential data is

called the cover image, while the stego image is the

resulting image with the conﬁdential data hidden. Re-

cent years have witnessed a proliferation of studies on

the science of steganography about information with

many proposed methods to increase the concealment

and security of cover images, which can be divided

548

Azooz, H., Ben Salah, K., Kherallah, M. and Naceur, M.

A Novel Image Steganography Method Based on Spatial Domain with War Strategy Optimization and Reed Solomon Model.

DOI: 10.5220/0012366800003636

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 16th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2024) - Volume 3, pages 548-557

ISBN: 978-989-758-680-4; ISSN: 2184-433X

into two main groups: spatial domain and frequency

domain techniques. In the spatial domain, hidden in-

formation is required by directly processing the pixel

values of the cover image. These can be implemented

easily and provide a large inclusion capacity; how-

ever, they are highly detectable using hidden analy-

sis techniques. On the other hand, frequency domain

techniques include embedding hidden information in

the parameters of a modiﬁed version of the cover im-

age. The complexity of implementing these solutions

increases the potential beneﬁt of increasing security

against discovery, but its drawbacks remain. It is

the delay in masking higher information against spa-

tial domain techniques. And because the efﬁciency

of any information cloaking method depends heav-

ily on choosing the embedding region in the middle

of the envelope intelligently and accurately, we pro-

posed an innovative approach that does not exist in

advance based on the use of war strategy optimiza-

tion algorithms and Reed Solomon model to correct

the errors of the secret messages extracted, which en-

hances the reliability of message recovery. This paper

explores in detail the information steganography tech-

nique that is used to replace the least signiﬁcant bit

inside the host image to include the bits of the secret

message and is adopted in choosing the embedding

points on the war strategy optimization and discusses

its theoretical foundations, implementation and exper-

imental results. The contribution of this research lies

in the advancement of secure information steganog-

raphy practices, which Provides valuable insights to

effectively hide conﬁdential data within images. The

efﬁciency of any cloaking method can be measured by

using geometric parameters PSNR, histogram, ssim

and BER, which we used to evaluate the stego image

quality produced by our proposed system.

2 RELATED WORKS

Steganography uses handwritten documents to hide a

secret message. Its secure communication and data

protection capabilities have drawn attention in re-

cent years. In examining a similar study (Ayyarao

et al., 2022), war strategy optimization, a new algo-

rithm inﬂuenced by war principles, was mentioned.

It solves difﬁcult optimization problems and creates

strong and safe data-hiding schemes when combined

with a masking algorithm. It strikes a new bal-

ance between exploration and exploitation. (Jaradat

et al., 2021) proposed a new steganography method

that uses chaotic partial swarm optimization (CPSO)

to achieve high embedding capacity. The cover im-

age and secret message are divided into blocks, and

each block stores an appropriate amount of secret bits.

Cops involve chaotic dynamics and optimization pro-

cesses. Conventional methods cost less computation-

ally than proposed ones. Steganography methods af-

fect embedding and extraction performance. In the

(Li and He, 2018) proposed employing pixel-value

differencing and PSO to hide critical data in the cover

image. The authors in the (Shah and Bichkar, 2018)

used a liner convergence generator and the genetic al-

gorithm (GA), they were able to embed secret infor-

mation into the cover image by specifying the appro-

priate locations to place it (using at least two bits per

pixel) the proposed model offered strong data clock-

ing at the expense of embedding capacity which was

reduced to just two bits. The authors of (Swain,

2019) used differencing and substitution mechanisms

to hide high-capacity information the LSB two bits

are substituted with zeros, and then the remaining

six bits undergo quotient value differencing (QVD).

In (Nipanikar et al., 2018), an embedding method

based on the use of PSO for optimal selection of

pixel and wavelet transformation with the goal of hid-

ing a secret sound signal in the cover image. In the

(Mohsin et al., 2019) propose a new technique for

image steganography based on PSO by using pixel se-

lection for the concealment of secret data and the spe-

cial domain where are used to ﬁnd the optimal pixel

in the cover image to embed the secret data based on

genetic algorithm. Despite introducing a novel ap-

proach for concealing images with a signiﬁcant em-

bedding capacity, the experimental outcomes of this

method did not yield a commendable peak signal-to-

noise ratio (PSNR) value. These ﬁndings were com-

paratively lower than those obtained using the genetic

algorithm (GA). (Sharma and Batra, 2021) Proposed.

PSO. Based on Hoffman’s encoding HE. Method for

image steganography. The results of the CI experi-

ment are. Discussed. As are the implications of us-

ing hidden messages of varying sizes. Although it

improved the performance and efﬁciency of informa-

tion steganography, it did not add visual quality val-

ues beyond the results of our proposed approach. Us-

ing particle swarm optimization (PSO), Muhuri et al.

(Muhuri et al., 2020) developed image steganogra-

phy on integer wavelet transformation (IWT) To lo-

cate the best possible pixel in which to conceal the

secret data within the cover image. To precisely. Lo-

cate the molten iron tanker. The authors employed

the grayscale image matching Techniques to evaluate

the cross marks on the Vessel Particle swarm analysis

is utilized to roughly determine the optimal matching

point of the picture and then they improved Harnis

corner detection algorithm and the sub-pixel approach

are employed for exact positioning in the process of a

A Novel Image Steganography Method Based on Spatial Domain with War Strategy Optimization and Reed Solomon Model

549

grayscale image matching analysis. The discrepancy

between errors was similarly diminished in the case

of the original integer wavelet. The coefﬁcient values

were determined and subsequently adjusted through

an optimal pixel adjustment procedure. Bedi et al

(Bedi et al., 2013) offer a spatial domain data hiding

approach using PSO to identify the optimal pixel lo-

cation for hiding one image within another image by

improving the SSIM index. this scheme’s stego image

quality was higher than that of the dynamic program-

ming and GA- based LSB schemes despite it is mod-

est embedding capacity. El Eman in (El-Emam, 2015)

developed. An Adaptive neural network-based modi-

ﬁcation to the PSO-based data concealing technique.

RS Codes are widely used in failure recovery of stor-

age systems (Tang and Zhang, 2021) where RS codes

are deﬁned using parity check matrices which are ei-

ther lined Cauchy matrices, Padded with an identity

with a fewer ‘1’. or van der Monde are used in the

deﬁnition of RS codes. Researchers are exploring

low-density parity check [LDPC] methods for mas-

sive data storage tang and zank demonstrate how a

vander Monde mat matrix can be joined to an identity

Matrix to ease the design of encoders and decoders.

Because of its widespread use, researchers have been

hard at work perfecting RS code Enhancing encoding

efﬁciency and inventing cutting-edge decoding meth-

ods (Gunjal and Sonawane, 2023). (Xu, 2022) Use

a steganography algorithm based on the least signiﬁ-

cant bits and RScode to code secret data before em-

bedding it in the carrier image and have been shown

to provide higher attack immunity with a slight cost to

imperceptibility and capacity. According to (Wolpert

and Macready, 1997) there is no single optimization

algorithm that gives ideal results for current optimiza-

tion problems. Because research is still ongoing in

this ﬁeld to discover new optimization algorithms for

hiding conﬁdential data, our proposed algorithm is

innovative; this paper proposes a new technique To

meta-optimize ﬁnding pixel-perfect embedding posi-

tions for secret message bits in media based on the

strategic warfare algorithm and RS codes. Where the

proposed system not only provides compatibility with

the RS model, higher capabilities for evaluating the

information, but also maintains the image quality. In

(Azooz et al., 2023) introduced a novel approach that

transforms the concealed message into a Novel En-

hanced Quantum Representation (NEQR) code, em-

ploying a quantum encoding framework for secrecy

and integrity. Placing the quantum circuit at K-means

algorithm-generated cluster centroids seamlessly con-

ceal the message within the cover image. The paper is

organized as follows. In the second section, an expla-

nation of the algorithm used, in the third section, the

results and experiments, and in the last section, the

summary.

3 PROPOSED METHODOLOGY

In our proposed approach, we used the War Strat-

egy Optimization algorithm to ﬁnd the best param-

eters for embedding the secret message to make the

stego image perceptibly similar to the original cover

image. The system includes four functions, initial-

ization, WSO, steganography objective function, and

embedding secret message. Algorithm (1)

3.1 Initialization

Initialization works on initializing the search agents’

positions, which are called (soldiers) by creating ran-

dom groups within a search space deﬁned between the

upper and lower bounds. The initialization function

creates a two-dimensional array of zeros, for exam-

ple (search-agents-no, dim), to store the positions of

soldiers who are search agents. In the context of our

proposed approach to hiding secret message bits, the

importance of the initialization function emerges be-

cause it ensures that search agents are initialized in a

way that enables them to ﬁnd the optimal solution in

determining the positions where secret message bits

will be hidden. The function is iterated across each

dimension of the search space dimensions; when the

upper- and lower-dimension arrays contain more than

one element, each factor along the dimension will be

assigned as a random value within its corresponding

bounds so that each position is within the search space

boundaries. However, if the upper and lower bound

arrays contain only one element, all factor positions

are assigned random values within their bounds.

3.2 War Strategy Optimization (WSO)

The algorithm is an independent, high-level optimiza-

tion algorithm that aims to provide strategies for de-

veloping algorithms to search for locations where se-

cret message bits are hidden. It uses a simulation of

ancient military strategies to track a group of peo-

ple, introduced as soldiers, whose task is to search for

solutions to an optimization problem. The essence

of this algorithm is to use successive algorithms to

solve complex optimization problems for which there

are no known effective algorithms. The innovative

WSO algorithm uses steganography to ﬁnd the best

values that control its behavior in terms of the num-

ber of search agents (soldiers), the maximum num-

ber of iterations, and the limits of the search space.

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

550

This determines how to include the secret message in

the cover image. Our proposed work aims to make

the WSO algorithm search for the best set of parame-

ters to create a stego image that is tactically similar to

the original cover image. In each iteration of the al-

gorithm, the positions of the search agents (soldiers)

are updated based on their suitability for the objective

location and the positions of other soldiers. This is

applied in the concealment algorithm by ﬁnding suit-

able positions for hiding secret message bits based

on pixel suitability for the position, in order to avoid

detection. The WSO algorithm starts by initializing

proposed steganography positions within the deﬁned

search space using upper and lower bounds, then eval-

uating the suitability of each hiding position using a

deﬁned objective function. The best position is cho-

sen to represent the king. The main loop of the WSO

algorithm starts in each iteration by choosing what is

called in the algorithm (the Viceroy or Co-King) to be

the second-best proposed hiding position after sorting

all search agent positions according to their suitability

for the mentioned position. Then a random number is

created for each soldier; if it is less than 0.1, that sol-

dier’s new position or location is calculated by mov-

ing it towards the best position between the king and

co-king in a direction away from the king. A random

number is generated for each soldier; if this number

is less than 0.1, then it is moved and placed in a new

position by moving it toward the position of the king

and co-king.

In the next step, this new position is checked

against upper and lower bounds to ensure that ob-

tained values are within these bounds; otherwise, it

is clipped outside these bounds. This new position’s

ﬁtness is then evaluated using a given objective func-

tion. To compare this new position’s ﬁtness value

with that of the current king, if it is better, then this

new position becomes that of the new king. However,

if this ﬁtness value is also better than those in this

soldier’s previous position but worse than that of the

current king, then this new soldier’s position replaces

its previous one. The accuracy and focus of this al-

gorithm on ﬁnding places to hide secret message bits

inside a cover image gives it an advantage over previ-

ous methods and algorithms used.

After all soldier positions are updated, if there

have been less than 1000 iterations, one soldier

(the one with the worst ﬁtness) has its repetition

counter increased and its position randomly reinitial-

ized within search space limits for preventing algo-

rithm initialization during the initialization process,

it does not get stuck in the local minimum, and this

leads to an increase in the iteration counter to keep

track of the number of iterations that have been made

so far, until all iterations are completed, the algorithm

returns the best solution that was found, which is (the

position of the king) along with its ﬁtness value.

3.3 Steganography Object Function

In the proposed system, we used an objective function

for the War Strategy Optimization (WSO) algorithm

to ﬁnd the optimal parameters for embedding a se-

cret message in a cover image by maximizing the per-

ceptual similarity between the two images. This step

calculates the perceptual similarity between the cover

and the stego images for a set of embedding parame-

ters. The objective concealment function converts the

similarity measure into a dissimilarity measure by re-

turning its negative value, allowing it to be minimized

by the WSO algorithm. With each iteration, the WSO

algorithm converges towards the optimal solution by

ﬁnding the set of embedding parameters that provide

the highest perceptual similarity between the original

and resulting images. After completing its work, the

algorithm returns the optimal parameters that can be

used to embed the secret message in the cover image.

The optimal parameters are those that result in the

highest perceptual similarity between the cover and

stego images. After running the WSO algorithm, it

represents its best output from the three as the king,

and the best ﬁtness value is chosen as (the king’s ﬁt-

ness), which is the highest perceptual similarity be-

tween the two images. The objective concealment

function extracts the number of least signiﬁcant bits

and sends them to the next embedding process for use

in embedding the secret message in the cover image.

In our proposed approach, the concealment improve-

ment function is integrated into the War Strategy Op-

timization algorithm. to extract correct value for pa-

rameter determining number of least signiﬁcant bits

from king matrix returned by WSO algorithm, then

passing it to secret message embedding process using

Reed-Solomon error correction code model to correct

resulting errors. This makes WSO algorithm also re-

sponsible for improving parameter for number of least

signiﬁcant bits and position of secret message in cover

image.

3.4 Embedding Secret Message

The embedding process involves inserting the cover

image, the secret message, and the parameter for de-

termining the least signiﬁcant bits required for use in

embedding. This embedding process uses this param-

eter to create a binary bit mask to process speciﬁc bits

of another binary number using bitwise operations.

Here, the bit mask ((2

− 1)) is used to clear the least

A Novel Image Steganography Method Based on Spatial Domain with War Strategy Optimization and Reed Solomon Model

551

signiﬁcant part of each pixel in the cover image ac-

cording to equation number.

[i] = I

[i] AND (2

− 1) + b[i] (1)

Where I

is the stego image, I

is the cover image, b[i]

is a representation of the binary string of the secret

message, n is the number of LSB, and [i] is the pixel.

Algorithm 1: Proposed WSO-LSB-RSCodec Algo-

rithm.

Input : search-agents no, dim, ub, lb

Output: positions

(1) INITIALIZATION() Initialize positions;

(2) for each dimension i do

(3) positions[i] ← [lb − i, ub − i];

(4) return positions;

(5) WSO(soldiers no, max-iter, Lb, Ub, Dim) Initialize

king and king-ﬁt;

(6) Initialize positions using

INITIALIZATION(search-agents no, dim, ub, lb);

(7) Initialize old ﬁtness and new ﬁtness;

(8) for each position do

(9) Compute ﬁtness using

STEGANOGRAPHY OBJECTIVE((steganography

objective) function);

(10) if ﬁtness < king-ﬁt then

(11) Update king-ﬁt and king with ﬁtness and

position, respectively;

(12) l = 0;

(13) while l < max-iter do

(14) Update positions and ﬁtness based on the

WSO algorithm rules;

(15) for each position do

(16) if ﬁtness < king-ﬁt then

(17) Update king-ﬁt and king with ﬁtness

and new position, respectively;

(18) if ﬁtness < ﬁtness-old then

(19) Update positions and ﬁtness-old with

new position and ﬁtness,

respectively;

(20) l+ = 1;

(21) return king-ﬁt, king, convergence-curve;

(22) STEGANOGRAPHY OBJECTIVE(cover image, secret

message) Embed the secret message into the

cover image using no LSBs to get stego image;

(23) Compute and return the negative SSIM between

cover image and stego image;

(24) EMBED MESSAGE(cover-image, secret-message)

Initialize RSCodec object for Reed-Solomon

codes;

(25) Encode secret-message using RSCodec;

(26) Convert encoded-message to binary string and

ﬂatten cover-image to 1D array;

(27) Embed binary message into LSBs of cover image

ﬂat to get stego image ﬂat;

(28) return stego image;

3.5 Reed-Solomon Codes

Reed-Solomon codes were introduced by Gustavus

Solomon and Irving S. Reed. They are a subclass of

non-binary BCH codes, but unlike binary encoders,

they operate on multiple bits at a time. The basic

principle of Reed-Solomon code operation is to re-

cover corrupted messages that are transmitted over

media such as the network and the Internet. It can

detect and correct errors that occur during transmis-

sion or even storage during the decryption process. In

our proposed approach, we used the RS Code Class

model to receive secret message data, detect errors

in it, and correct them. Then, we encrypted it be-

fore embedding it in the cover image and used it later

to decrypt it after extracting it from the Stego im-

age. We chose the RS Code Class because it provides

a high-level interface for encrypting and decrypting

messages using Reed-Solomon codes. When creat-

ing an instance of the class, it determines the num-

ber of error-correction symbols to be used as parame-

ters and returns the message with the error-correction

codes appended. In the extraction process, it takes

the encrypted message and then returns a set contain-

ing the corrected errors and the extracted secret mes-

sage. The Reed-Solomon model relies on polynomial

fulﬁllment on ﬁnite ﬁelds, which is a set of elements

with addition and multiplication operations and con-

tains a limited number of elements. In the process of

encoding the secret message, the message codes are

assigned to elements of a ﬁeld of a speciﬁed size, and

a polynomial of degree (k − 1) on the speciﬁed ﬁeld is

created so that the message codes, which are polyno-

mial coefﬁcients, are evaluated at some of the charac-

teristic points to generate a codeword. This codeword

includes both secret message codes and veriﬁcation

codes. This is done by assigning a one-to-one map-

ping between the set of possible message codes and

the elements of the speciﬁed ﬁeld. The encoding pro-

cess for RS codes involves representing the message

to be encoded as a polynomial over a ﬁnite ﬁeld. As-

suming that the message to be encoded is represented

by the vector of symbols m = (m

, m

, . . . , m

k−1

), the

message polynomial p(x) is then deﬁned as:

p(x) = m

+ m

x + m

+ . . . + m

k−1

The codeword is generated by evaluating the message

polynomial at n distinct points a

, a

, . . . , a

n−1

in the

ﬁnite ﬁeld to obtain the codeword symbols c

= p(a

)

for i = 0, 1, . . . , n − 1. The ﬁrst k codeword symbols

are the original message symbols, and the remaining

n − k symbols are the parity check symbols.

In other words, RSCode is speciﬁed as RS(n,k),

where n is the length of the codeword, and k is the

dimension of the code extraction.

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

552

Figure 1: The stego document image quality measured by

PSNR and SSIM on Type 8 dataset.

3.6 Extraction Process

The secret message was extracted from a stego image

using the extraction algorithm, represented in Figure

1. The algorithm starts by ﬂattening the stego im-

age into a one-dimensional array, where the ﬂattening

method returns a new one-dimensional collapsed ar-

ray containing all the elements of the original array.

Flattening the image makes it easy to iterate over its

pixel values in the following steps.

In the next step, an empty string is initialized to

store the binary representation of the secret message.

In the third step, the function then enters a while loop

that iterates over the stego image to extract eight bits

from the image and take the least value of them and

link them in a sequential string (byte). After the loop

is ﬁnished, the secret message is represented as a se-

quence of bits b

, b

, . . . , b

, where n is the length

of the binary message in bits. These bits are assem-

bled into parts consisting of eight bits, and each part is

represented as an integer C

using the following equa-

tion:

∑

j=0

8i+ j

· 2

7− j

Where i is the index of the chunk, and j is the in-

dex variable. In the ﬁnal step, the parts represented by

integers are converted to their corresponding ASCII

character and then linked in a string and reshaped into

the original secret message as an extracted secret mes-

sage.

4 EXPERIMENTAL RESULTS

4.1 The Parameters

Table (1) presents the parameters used in our pro-

posed system for the WSO algorithm and the

RSCodes class of the Reed-Solo Model.

Algorithm 2: Proposed Extraction Algorithm.

Input : Stego-Image

Output: Secret-Message

1 Flatten the Stego-Image into a 1D array;

2 Initialize an empty binary string to store the

extracted binary message;

3 Set the extraction index to 0;

4 while extraction index < length of

Stego-Image do

5 Extract 8 bits from the Stego-Image by

extracting the least signiﬁcant bit of

each pixel;

6 if extracted bits form a marker indicating

the end of the secret message then

7 break;

8 Append the extracted bits to the binary

string;

9 Increment the extraction index by 8;

10 Convert the binary string into a string by

grouping its bits into 8-bit chunks and

converting each chunk to its corresponding

ASCII character;

11 return the extracted secret message;

4.2 Datasets

Due to the lack of databases used in the latest informa-

tion steganography evaluation techniques for image

documents, we have to examine the robustness of the

proposed system with a limited number of datasets.

We used eight sets of image documents, handwritten

document images, and standard grayscale images de-

scribed as type (1), type (2), type (3), type (4), type

(5), type (6), type (7) and type (8). The proposed sys-

tem was tested on 15 images of the ﬁrst and second

types, 12 images of the third and fourth type, 10 im-

ages of the ﬁfth type, and 20 images of the sixth and

seventh types, and 12 images of Type 8th and 10 im-

ages of type 9th after the experiment in (Jaradat et al.,

2021; Sharma and Batra, 2021).

4.3 Evaluation Metrics

Stego image ﬁdelity refers to the quality of an image

after embedding a conﬁdential message into a cover

document image. The quality of the stego image can

be measured using common stego document image ﬁ-

delity techniques, as deﬁned by Equation (9), with the

Peak Signal-to-Noise Ratio (PSNR). PSNR quantiﬁes

the ﬁdelity of the stego image by calculating the mean

square error (MSE), which represents the squared dif-

ference between the stego image and the cover im-

age. Enhanced ﬁdelity results from a higher PSNR,

A Novel Image Steganography Method Based on Spatial Domain with War Strategy Optimization and Reed Solomon Model

553

Table 1: Parameters Used in the WSO Algorithm and Reed-Solomon Model.

Method Parameters Designations Values

WSO Search agents (Soldiers) n - number of search agents 10

Max-iter Maximum number of iterations 500

lb Lower bound of search space [1]

ub Upper bounds of search space [8]

dim Dimensionality of the search space 1

fobj Objective function λx.x

Reed-Solomon Model n - total symbols in codeword 10

k - message data symbols -

indicating minimal distortion during the embedding

process.

The PSNR is calculated using the following for-

mula:

PSNR = 10 · log



255

MSE



(2)

Additionally, the Structural Similarity Index

(SSIM) is another powerful metric used to assess the

quality of a stego image. It compares the stego docu-

ment image to the cover image and provides a value in

the range of [0, 1], with values closer to 1 indicating

a higher degree of ﬁdelity.

The SSIM is computed using the following equa-

tion:

SSIM(x, y) =

(2µ

+ c

)(2σ

+ c

)

(µ

+ µ

+ c

)(σ

+ σ

+ c

)

(3)

In this equation, µ

and µ

represent the averages

of the cover and stego images, respectively. σ

rep-

resents the covariance between the cover and stego

images. Standard deviations are denoted by σ

and

. The parameters c

and c

are constants used to

stabilize the division and avoid division by zero.

SSIM takes into account factors such as average

intensity, contrast, and structure between the two im-

ages and is a valuable tool for assessing the quality of

stego images.

4.4 Quantitative Study

To justify the utilization of secret messages for docu-

ment steganography, our method was evaluated using

a speciﬁc dataset designed for this purpose. Among

the dataset types, Type 5 demonstrated superior per-

formance in terms of Peak Signal-to-Noise Ratio

(PSNR) compared to other datasets.

We employed two techniques for concealing

conﬁdential data within images of documents and

handwritten manuscripts. These techniques involve

data concealment both without employing the Reed-

Solomon model and with it. The utilized secret data

size was 144 bits. The results of applying embed-

ding position optimization using the War Strategy Op-

timization algorithm and the Reed-Solomon model

were showcased on eight data sets. The results of ex-

ecuting our proposed system on the utilized dataset

types are showcased in Table 2. This table presents

the average PSNR values for selected types of doc-

uments, handwritten manuscripts, and grayscale gra-

dient images. Furthermore, it compares these results

with classical-LSB and the related works. The results

in the table substantiate that our employed method ex-

hibits higher efﬁciency when compared to classical-

LSB and relevant works in terms of distortion mea-

surement metrics.

In Table 3, since no related works have previ-

ously used optimization algorithms on the dataset of

document images and handwritten manuscripts, we

conducted a more speciﬁc comparative study on nine

grayscale images with related works. From this, we

deduced that our proposed approach for concealing

conﬁdential data using the Least Signiﬁcant Bit (LSB)

method along with the War Strategy Optimization al-

gorithm and Reed-Solomon coding is more efﬁcient

in terms of quality compared to the existing enhance-

ments in relevant research.

Based on the results obtained using both PSNR

and SSIM measures in Table 3, our proposed system

achieved better stego image quality compared to clas-

sical LSB and Muhuri, Pranab K et al and Sharma et

al approachs.

4.5 Qualitative Study

The steganography technique is secure if it is resis-

tant to various steganography analysis attacks. The

security of the information hiding technique used in

our proposed system can be evaluated through pixel

difference histogram analysis and the Human Visual

System (HVS). in Figure (4). It can be noted that the

graph in the stego image is very similar to the graph

of the cover image, which is an indicator of the ef-

fectiveness of our hiding system. To use histogram

analysis of pixel differences between the cover image

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

554

Table 2: Average PSNR Values and Comparisons.

Benchmark Number of Documents PSNR WSO-LSB WSO-LSB-RSCodes C-LSB

Type 5 10 85.71 87.84 55.24 -

Type 6 20 87.01 87.13 54.12 -

Type 8 12 84.85 86.87 56.15 63.25

Type 9 10 85.32 87.69 55.27 -

Table 3: PSNR and SSIM Comparisons.

Image Our

System

without

RS Codes

Our Sys-

tem with

RS Codes

Classical

LSB

Jaradat

et al.

(2021)

Nipanikar

et al.

(2018)

Mohsin et

al. (2019)

Sharma

et al.

(2021)

Baboon 80.62 81.82 51.14 46.65 78.65 0.999 0.999

Lena 81.87 87.69 51.13 51.64 78.65 0.999 0.9981

Airplane 80.77 81.94 51.14 51.35 55.24 63.41 70.35

Boat 84.85 86.87 51.13 51.42 70.92 75.31 80.24

Pepper 80.71 81.82 56.15 59.31 66.43 70.82 75.75

Barbara 85.05 87.00 51.14 51.13 60.21 64.61 69.54

Couple 83.05 86.08 51.12 51.39 71.82 76.21 81.14

Jet 80.77 81.94 51.14 51.35 65.12 69.51 74.44

Goldhill 80.86 81.86 51.14 50.63 70.95 75.35 80.28

Figure 2: Visual comparison of PSNR between WSO-LSB-

RSCodes system and related works on 9 images of Type 8

dataset.

Figure 3: The stego document image quality measured by

PSNR and SSIM on Type 8 dataset.

and the stego image. Figure No.3 shows a compari-

son between SSIM and PSNR values for our proposed

system. The same image from dataset type 9 used in

stego image quality tests were used to prove that the

proposed method produces average PSNR and SSIM

values close to (1). Therefore, we conclude that the

resulting stego document image from our proposed

WSO-LSB-RScode system is of high quality.

5 ROBUSTNESS OF THE

RETRIEVED SECRET

MESSAGE

Table 4: Accuracy Ratio of Documents Embedded.

Benchmarks Name PSNR SSIM BER

Type 2 MNIST 86.87 0.99 100

Type 4 CASIA 86.87 0.99 100

Type 6 Tobacco800 87.13 0.99 100

Type 7 L3iDocCopies 80.98 0.99 100

Type 8 Std. Grey Image 87.00 0.99 100

Type 9 Dataset-1 87.69 0.99 100

The Bit Error Rate (BER) was used to validate the

reliability of the extracted encrypted message. In our

proposed system, the bit error rate is calculated using

the following equation:

We conducted an evaluation of the robustness of

the WSO-LSB-RSCodes system against noise by in-

troducing varying levels of noise to the stego image

for three different types of noise: salt and pepper,

Gaussian, and speckle noise as shown in Table 4. We

then extracted the secret message from these images,

compared it with the original message, and calculated

the Bit Error Rate (BER) for each dataset type used.

The results we obtained indicate that the data hid-

ing technique used in WSO-LSB-RSCodes is highly

resilient against noise, even when the document is ex-

posed to noise multiple times as needed during its us-

age. The technique employed in our data hiding ap-

A Novel Image Steganography Method Based on Spatial Domain with War Strategy Optimization and Reed Solomon Model

555

Table 5: Benchmark: Pixel Differences Histogram Analysis for Cover and Stego Images.

Type Cover image Stego image Pixel Differences

2 C2 S2 0.0001

7 C7 S7 0.0005

8 (couple) C8 S8 0.0001

9 C9 S9 0.0001

Figure 4: Comparison of Histogram Visualization for Cover

and Encoded Images.

proach has proven its effectiveness, even against chal-

lenging types of noise such as speckle noise.

We evaluate stego images against JPEG compres-

sion attack using relatively low quality to test the

quality of the visual image. The results obtained

by the Accuracy Ratio (AccR), and the relation-

ship between the imperceptibility and capacity of the

steganography protocol by testing benchmarks (Type

1, 2, 3, 5, 6, 7, 8) indicate that there are no errors

in extracting hidden data from the image after stego.

It underwent lossy JPEG compression, which leads

to the conclusion that the approach used has a better

ability to withstand compression and distortion com-

pared to related work.

Filtering is a process of applying a ﬁlter to images

to mitigate or change pixel values in a stego image and

make the extraction of the secret message more difﬁ-

cult. We used ﬁltering in a ﬁltering attack on the stego

image using ﬁlters (Gaussian ﬁltering, Median ﬁlter-

ing, and Bilateral ﬁltering), and then extracted the se-

cret message from the image that had been ﬁltered.

Testing was done using the BER metric on dataset

groups (Type 1, 2, 3, 5, 6, 7, 8) for each ﬁltering at-

tack, and the test results were 100% for all datasets.

A decrease in the BER index indicates better strength

of the stego image against ﬁltering and is an evalu-

ation factor for the success of the extraction process

and the quality of the extracted secret message. As

can be seen from the bit error rate values, Median ﬁl-

tering has the least impact on the extracted message,

which indicates that our proposed system retains the

integrity of the secret message better even when ex-

posed to Median ﬁltering.

6 CONCLUSION

A novel steganography system for document and

handwriting images is proposed, leveraging the War

Strategy Optimization (WSO) algorithm. This inte-

gration enables a dynamic approach to ﬁnding opti-

mal embedment positions, addressing the challenge

of enhancing steganography methods. The system fo-

cuses on maximizing hidden data while minimizing

the impact on cover data, bolstering security against

steganalysis. By combining optimization techniques

with perceptual similarity measures, the system en-

hances security and robustness in image steganogra-

phy. Reed-Solomon error-correcting codes add an ex-

tra layer of reliability, improving data retrieval accu-

racy and minimizing quality decline. Experimental

results, based on quality metrics (PSNR, SSIM, BER,

HVS), highlight the high quality of stego images. On-

going research aims to extend system capabilities, in-

tegrating steganographic and cryptographic methods

for manuscripts’ protection and enhancing effective-

ness across diverse cover media and environments.

REFERENCES

Ayyarao, T. S. et al. (2022). War strategy optimization al-

gorithm: a new effective metaheuristic algorithm for

global optimization. IEEE Access, 10:25073–25105.

ICAART 2024 - 16th International Conference on Agents and Artiﬁcial Intelligence

556

Azooz, H. J., Salah, K. B., Kherallah, M., et al. (2023).

Quantum steganography: Hiding secret messages in

images using quantum circuits and sift. International

Journal of Advanced Computer Science and Applica-

tions, 14(10).

Bedi, P., Bansal, R., and Sehgal, P. (2013). Using pso in

a spatial domain based image hiding scheme with dis-

tortion tolerance. Computers and Electrical Engineer-

ing, 39(2):640–654.

El-Emam, N. N. (2015). New data-hiding algorithm based

on adaptive neural networks with modiﬁed particle

swarm optimization. Computers and Security, 55:21–

45.

Gunjal, M. B. and Sonawane, V. R. (2023). Cloud data re-

covery and secure data storage with advanced cauchy

reed-solomon codes and role-based cryptographic ac-

cess. Scandinavian Journal of Information Systems,

35(1):859–869.

Jaradat, A., Taqieddin, E., and Mowaﬁ, M. (2021). A high-

capacity image steganography method using chaotic

particle swarm optimization. Security and Communi-

cation Networks, 2021:1–11.

Li, Z. and He, Y. (2018). Steganography with pixel-

value differencing and modulus function based on

pso. Journal of Information Security and Applica-

tions, 43:47–52.

Mohsin, A. H., Zaidan, A., Zaidan, B., et al. (2019). New

method of image steganography based on particle

swarm optimization algorithm in spatial domain for

high embedding capacity. IEEE Access, 7:168994–

169010.

Muhuri, P. K., Ashraf, Z., and Goel, S. (2020). A novel im-

age steganographic method based on integer wavelet

transformation and particle swarm optimization. Ap-

plied Soft Computing, 92:106257.

Nipanikar, S. I., Hima Deepthi, V., and Kulkarni, N. (2018).

A sparse representation based image steganography

using particle swarm optimization and wavelet trans-

form. Alexandria Engineering Journal, 57(4):2343–

2356.

Shah, P. D. and Bichkar, R. S. (2018). A secure spa-

tial domain image steganography using genetic algo-

rithm and linear congruential generator. In Interna-

tional Conference on Intelligent Computing and Ap-

plications, pages 119–129.

Sharma, N. and Batra, U. (2021). An enhanced huffman-

pso based image optimization algorithm for image

steganography. Genetic Programming and Evolvable

Machines, 22:189–205.

Swain, G. (2019). Very high capacity image steganogra-

phy technique using quotient value differencing and

lsb substitution. Security and Communication Net-

works Arabian Journal for Science and Engineering,

44(4):2995–3004.

Tang, Y. J. and Zhang, X. (2021). Fast en/decoding of reed-

solomon codes for failure recovery. IEEE Transac-

tions on Computers, 71(3):724–735.

Wolpert, D. H. and Macready, W. G. (1997). No free lunch

theorems for optimization. IEEE transactions on evo-

lutionary computation, 1(1):67–82.

Xu, J. (2022). Improved least signiﬁcant bit algorithm based

on rs-code. In Second International Conference on

Advanced Algorithms and Signal Image Processing

(AASIP 2022), volume 12475.

A Novel Image Steganography Method Based on Spatial Domain with War Strategy Optimization and Reed Solomon Model

557