SECURING CONFIDENTIAL MESSAGES USING THE IRSA-PI-PVD MODEL IN

THE SPATIAL DOMAIN OF IMAGE STEGANOGRAPHY

Karthikeyan Na*, Shanthi Sb

a
Assistant Professor, Department of Computer Science and Engineering, Government College of
Engineering, Sengipatti, Thanjavur, Tamil Nadu, India
b
Associate Professor, Department of Computer Science and Engineering, Kongu Engineering College,
Perundurai, Erode, Tamil Nadu, India

*Corresponding Author:
Mr. N. Karthikeyan,
Assistant Professor,
Department of Computer Science and Engineering, Government
College of Engineering, Sengipatti, Thanjavur, Tamil Nadu, India
[email protected]

1
SECURING CONFIDENTIAL MESSAGES USING THE IRSA-PI-PVD MODEL IN
THE SPATIAL DOMAIN OF IMAGE STEGANOGRAPHY

ABSTRACT

The rapid expansion of recent technologies has led to a significant increase in network
usage. Safeguarding the privacy of data has emerged as a crucial consideration in the
exchange of confidential information during communication. In order to enhance the
protection of such sensitive data, diverse methodologies have been formulated by various
researchers, including cryptography, steganography, and watermarking. Within this context,
the author presents an all-encompassing four-layer security model that integrates
cryptography and steganography techniques to reinforce the protection of confidential
information in the spatial domain of image steganography. The author introduces the IRSA-
PI-PVD model [ Improved RSA (IRSA) -Pixel Indicator (PI) -Pixel Value Differencing
(PVD)] as a mechanism for ensuring data security. The initial phase entails the encryption of
confidential data using the IRSA algorithm, followed by the reversal of the binary
representation of each encrypted character in the concealed message. Subsequently, the
reversed encrypted binary bits are seamlessly incorporated into the cover image using the
PVD technique, with the pixel indicator functioning as a reference channel for the
embedding process. The principal objective of this proposed model is to increase the
embedding capacity while simultaneously enhancing the image quality and mitigating
various vulnerability attacks.

Keywords: Steganography; Improved RSA; Pixel Indicator; Pixel Value Difference;

Subject Index: CO17

INTRODUCTION

In the realm of secure communication and data protection, the fields of cryptography and

steganography play pivotal roles. Cryptography involves the art and science of designing techniques to

secure information through encryption, ensuring that only authorized recipients can decipher the intended

content. It has evolved to encompass various encryption algorithms and protocols, striving to withstand

2
the challenges posed by increasingly sophisticated cyber threats. On the other hand, steganography

involves the subtle concealment of information within seemingly innocuous carriers, such as images,

audio, or video files. Its aim is to achieve confidentiality by hiding the very existence of the

communication. Steganography has expanded its horizons to leverage the vast digital media landscape,

employing innovative methods to embed information covertly while evading detection. The convergence

of cryptography and steganography has led to a new paradigm of secure communication, where the

strengths of each field are leveraged to fortify the weaknesses of the other. By combining cryptographic

techniques with steganographic concealment, a multi-layered security approach emerges, ensuring not

only the confidentiality of the message but also the very secrecy of its transmission.

The devised model implements a comprehensive four-layer security mechanism for safeguarding

confidential messages within the spatial domain of image steganography (Abdulhammed and Omar

Younis 2023). The process commences by encrypting the message using the IRSA algorithm (Nasution et.

al. 2020; Amalarethinam et. al. 2017) followed by the reversal of the binary representation of the

encrypted character bits. Subsequently, the reversed encrypted secret bits are seamlessly integrated into

the cover image through the utilization of the PVD technique (Zhang, Xinpeng, and Shuozhong Wang

2004; Wu, Da-Chun and Wen-Hsiang Tsai 2003; Swain and Gandharba 2015; H –C et al. 2005,

Gandharba Swain 2019). This embedding is performed by employing the pixel indicator as a reference

point, thereby incorporating the details into the additional channels of the same pixel. Consequently,

access to the concealed information is granted solely to authorized users possessing the knowledge of the

precise location of the secret bits within the stego image.

The key contributions of this research are summarized as follows. First, the proposed model introduces

a novel data encryption approach designed for secure communication and verification by leveraging

distinct asymmetric encryption techniques, which enhances both confidentiality and integrity in data

transmission. Second, a comprehensive performance evaluation is conducted, examining multiple

3
dimensions such as processing time, message capacity, and the strength of security mechanisms. This

analysis provides a detailed understanding of the system's efficiency and resilience against potential

threats. Finally, the proposed encryption model is compared with existing approaches using a range of

indicators, including time efficiency, capacity optimization, the quality and preservation of the cover

image, and overall security performance. This comparative analysis highlights the advantages of the new

model in terms of both operational efficiency and protection against vulnerabilities.

LITERATURE SURVEY
(Kaku Lishomwaa and Aaron Zimbab 2023) introduce an innovative strategy to enhance the security of

medical diagnosis records. The approach employs AES – 128 algorithms for encryption, utilizes the Diffie

Hellman key Exchange algorithm for secure key exchange, and applies the LSB technique to seamlessly

integrate a confidential message within a cover image. This comprehensive technique safeguards sensitive

medical information throughout its transmission and storage, bolstering confidentiality. By incorporating

encrypted image steganography methods, the research addresses the prevention of unauthorized access to

concealed medical data. This study sheds light on the critical importance of preserving privacy in medical

data management, offering potential advancements in the secure handling of medical records.

(Supriadi Rustad et. al. 2022) presented an innovative approach that employs a distinct utilization of

the inverted Least Significant Bit (LSB) within pixel values, for the purpose of encoding confidential data

into images. By employing an adaptive pattern, their technique strives to enhance the inconspicuousness

of the concealed information, thereby complicating unauthorized detection of the embedded data. The

experimental outcomes effectively demonstrate the inconspicuous nature of the inverted LSB image

steganography method, achieving a perceptibility score of 57.1324 and showcasing a maximum data

embedding capacity of 65536 bits.

(Dharwadkar et. al. 2023) introduced a novel steganography scheme for medical images that effectively

tackles the falling-off boundary concern by utilizing the PVD Method. This inventive strategy notably
4
amplifies the embedding capacity while concurrently upholding imperceptibility in medical imagery.

Through adept utilization of discrepancies in pixel values, the scheme adeptly conceals a substantial

volume of data within the images, successfully surmounting the boundary obstacle. This technique

demonstrates considerable potential for secure transmission of medical data, boasting a noteworthy

average embedding capacity of 865,423 bits alongside a crucial imperceptibility factor of 34.29dB,

thereby ensuring the preservation of data confidentiality and integrity.

(Monalisa Sahu et al. 2022) propose a model to enhance multi-directional Pixel-Value Differencing

(PVD) steganography to improve the embedding capacity. It is mainly focused on solving issues such as

Peak Difference Histogram (PDH) problems and boundary challenges commonly encountered in

conventional PVD techniques. The model investigates the advantages of their approach in improving the

quality of stego-images and bolstering their resilience against various attacks. The model they present

achieves an average Peak Signal-to-Noise Ratio (PSNR) of 34.89 and possesses a maximum embedding

capacity of 728,921 bits.

(Kalita et. al 2019) propose an innovative technique termed adaptive color image steganography, which

merges adjacent PVD with the LSB substitution method. This fusion maximizes the potency of image

steganography by capitalizing on the synergistic potential of these two approaches. Through meticulous

analysis and experimentation, the study underscores the effectiveness of their strategy in accomplishing

durable and imperceptible data embedding. The integration of adjacent pixel value differencing augments

security and proficient data concealment, while the incorporation of LSB substitution further bolsters the

steganographic procedure. The technique offers an average embedding capacity of approximately 3 bpp, a

PSNR spanning 38.1840, and a data embedding potential of 640,913 bits.

(Alexander et al. 2019) present a novel irreversible spatial domain data hiding method denoted as

A2MLSB. This technique is designed to adapt to both pixel value (PV) and pixel value difference (PVD)

by employing variable thresholds within their respective ranges. Additionally, the method utilizes a

5
modified least-significant bit (MLSB) substitution strategy, tailored specifically for grayscale images. The

proposed method boasts an average embedding capacity of 543,623 bits, accompanied by an average

PSNR value of 32.49 dB, demonstrating its capability to effectively resist various potential cyber attacks.

(Tseng et. al. 2015) introduce an innovative method that ingeniously incorporates pixel-value

differencing (PVD) along with a reference derived from a quantization table linked to perfect square

numbers, aiming to embed secret bits. This innovation significantly bolsters the effectiveness and security

of data concealment within digital images. Demonstrating substantial potential, the method supports the

embedding of 194,391 bits, maintaining an average imperceptibility of 46.83 dB. Furthermore, the

proposed technique showcases enhanced resilience against a spectrum of steganography cyberattacks,

marking a substantial advancement in the realm of secure data hiding.

PROPOSED SYSTEM
The proposed model employs a four-layer security approach to conceal secret messages within the

spatial domain of a cover image. Figure 1 illustrates the conceptual workflow of the proposed model for

both the embedding and extraction processes. To begin the embedding process, the user encrypts the

secret message using an improved RSA algorithm and then converts it into a binary form. Each converted

character's binary representation is reversed. Subsequently, the user analyzes the various color

components of each pixel in the cover image based on their highest intensity count to identify the most

dominant color component, which serves as a pixel indicator. This pixel indicator is used to determine

which channels of the same pixel will be used for embedding (Thanikaiselvan et. al. 2014).

6
 Fig.1 Conceptual Framework of the proposed model

Table 1 Range Quantization Table

Range 0-7 Aug-15 16-31 32-63 64-127

No. of Bits 3 3 4 5 6

Table 2 Possible pixel indicator values and actions

Pixel Indicator Channel1 Channel2

0 Embedding not done Embedding not done
1 Embedding done Embedding not done
10 Embedding not done Embedding done
11 Embedding done Embedding done

The user then examines each component of the pixel to check for a falling-off condition. If no falling-

off condition occurs, the user calculates the number of bits say ‘n’ that can be embedded into the specific

component of the pixel using a range quantization table represented in Table 1. The next 'n' possible bits

from the reversed encrypted binary message are extracted and embedded into the cover image using the

PVD technique. This embedding process is repeated for the other two components of the pixel, excluding

7
the dominant one, until all the reversed encrypted binary bits are embedded. To maintain the updated

details of the other two components of the particular pixel, the pixel indicator is modified using the PVD

technique, utilizing values from the possible pixel indicator value table is shown in Table 2. The complete

procedure for implementing the proposed model during the embedding process is outlined below.

a. Begin by prompting the user to input the secret message (M) and then select a suitable cover image

that will serve as the medium for embedding this confidential data.

b. Randomly generate four distinct large prime numbers, denoted as 𝑝,s.

c. Calculate the value of N = p*q*r*s

d. Compute totient function ɸ(g) = (p-1) * (q-1) * (r-1) * (s-1)

e. Select the random co-prime number e for ɸ(g), where GCD(e, ɸ(g) ) =1 and e1>1

f. Determine the unique private key d such that 𝑑*e=1 mod ɸ(g).

g. Encrypt the secret message using the expression C=Me mod N

h. Convert the encrypted secret message into its binary representation and then perform a reversal of

this binary sequence.

i. Examine the cover image to identify the dominant color component by assessing the frequency

count of each color channel in every pixel.

j. Embed the bits of the reversed encrypted message into the two remaining color channels of each

pixel in the cover image, using the dominant color channel as a reference.

k. Continue the embedding process iteratively until every bit of the reversed encrypted message has

been successfully incorporated into the cover image.

In the extraction process, the user selects the required stego image as input. The user then extracts the

length of the embedded message, private and public key pairs, and pixel indicators from the stego image

using the PVD technique. Based on the length of the embedded message, the user extracts the secret

messages from the stego image using the details of the pixel indicator for each pixel. The secret message

8
bits are extracted using the PVD technique. After extracting all the secret message bits, the bits are

grouped based on the key size and converted into an equivalent decimal representation. Finally, improved

RSA is applied for the decryption process, converting the extracted value into its equivalent character

representation.

While chaotic-based encryption is computationally faster, it is primarily suited for lightweight

applications and tends to be highly sensitive to initial conditions, parameter settings, and numerical

precision. This can introduce potential vulnerabilities if not carefully controlled. In contrast, the IRSA

algorithm is more versatile and efficient across a broad range of hardware and software platforms. IRSA

has undergone extensive testing, demonstrating strong resistance to cryptographic attacks, with its security

further enhanced by increasing key sizes. As a result, RSA remains a more reliable, interoperable, and

trusted solution for a wide variety of applications. The complete procedure for implementing the proposed

model during the extraction process is outlined below.

a. The extraction process begins by selecting the appropriate stego image from either a private

cloud server or a transmission channel, ensuring that it contains the embedded hidden message.

b. The PVD technique is applied to extract the reversed encrypted binary bits from the various color

channels of each pixel, using the dominant color channel as a reference.

c. After the binary bits are extracted, a binary reversal operation is performed to reverse the bits and

retrieve the encrypted secret data.

d. The reversed binary bits are grouped according to the specified key size, converted into their

corresponding decimal values, and the decrypted message (M) is then obtained using the

expression 𝑀=𝐶𝑑 mod 𝑁.

e. Finally, the decrypted values are converted into characters and combined to reconstruct the

original secret message that was concealed within the stego image.

9
 The PI-PVD method is presented as a significant advancement in the realm of image steganography,

surpassing the simple implementation of existing techniques. This approach sets itself apart from

traditional PVD by utilizing a pixel indicator to identify the most suitable pixels for data embedding,

thereby improving both the efficiency and quality of the embedded data. In contrast to conventional PVD,

which assesses embedding capacity based on the differences between adjacent pixel values often resulting

in visible distortions PI-PVD enables a more accurate and controlled embedding process. As a result, it

achieves higher data capacity and enhanced image quality, effectively minimizing artifacts while

preserving the visual integrity of the cover image. Overall, the introduction of PI-PVD represents a

noteworthy contribution to the field, offering improved performance metrics, including increased

embedding capacity and reduced distortion rates compared to standard PVD methods.

ASSESSMENT OF QUALITY
The proposed model considers quality, embedding capacity, and security, allowing evaluation of its

performance.

1.1. Assessment of Quality of the Cover image

The performance of the proposed system is justified through the assessment of its quality, which

measures the percentage of deviation of hidden information between the stego image and the cover image.

Evaluating the quality of the stego image involves using various quality metrics, and the details of these

metrics are provided below. The quality assessment takes inputs such as the Cover Image, Stego image,

and the dimensions of the image (PxQ).

1.1.1. Mean Squared Error (MSE)

MSE is utilized to assess the stego image's quality by measuring the average squared difference

between the pixel values of the cover image and the corresponding pixel values in the stego image. A

lower MSE value indicates a higher similarity between the cover and stego images, reflecting better image
10
quality with less distortion caused by the steganographic process. The calculation of MSE is defined in

Equation (1), where C[a][b][r] denotes the intensity value of the specific color channel in each pixel of

the cover image, and S[a][b][r] represents the intensity value of the corresponding color channel in each

pixel of the stego image.

P Q 3
1
∑ ∑ ∑
2
MSE = ( Cover [ a ] [b] [ k ] - Stego[a][b][k] ) (1)
3PQ a =0 b=0 k=0

1.1.2. Root Mean Squared Error (RMSE)

RMSE measures the difference between predicted values and actual values. It calculates as the square

root of the average of the squared differences between corresponding pixel values in the cover and stego

images. Lower RMSE values indicate a higher level of similarity between the images, reflecting better

image quality with less distortion introduced during the steganography process. The expression for the

calculation of RMSE is defined in Equation (2).

RMSE= √2 MSE (2)

1.1.3. Peak Signal to Noise Ratio (PSNR)

PSNR is utilized to measure the ratio between the maximum possible pixel value (peak signal) and the

MSE between the cover and stego images. It provides a quantifiable indication of how well the stego

image preserves the original image's quality. The expression for the calculation of PSNR is defined in

Equation (3).
2
PSNR = 10 log 10 ( MAX / MSE) (3)

1.1.4. Structural Context (SC)

The structural context plays a vital role in how the human visual system perceives the image's overall

structure and content, encompassing the visual details and connections among adjacent pixels. The

expression for the calculation of SC is defined in Equation (4).

11
 P Q 3

∑ ∑ ∑ ( Cover [ a ] [b ] [ k ])
2

SC = a=0
P
b=0 k=0
Q 3
(4)
∑ ∑ ∑ ( Stego[a][b][k]) 2

a=0 b =0 k=0

1.1.5. Structural Similarity Index (SSIM)

SSIM is a widely used image quality assessment metric that measures the similarity between two

images based on luminance, contrast, and structure. The expression for the calculation of SSIM is defined

in Equation (5), where µ c represents the average of the cover image, µ s represents the average of the

stego image, σc represents the variance of the cover image, and σ s represents the variance of the stego

image. Additionally, C1 and C2 are stabilization parameters, and for this model, the values considered are

C1 = 6.5025 and C2 = 58.25, respectively.

( 2 µc + us +C 1 ) (2 σ c +σ s+C 2)
SSIM = (5)
( µc 2+ µ s2+C 1 ) +(σ c 2+ σ s2 +C 2)

RESULTS AND DISCUSSION

The IRSA-PI-PVD model was successfully developed using Python within the Anaconda IDE

version 2023.07, operating on the Windows 10 platform. This development environment provides a robust

framework for efficiently implementing and testing the model. For the experimental analysis, a 24-bit key

size was used to represent each character in the hidden message. This choice of key size ensures a

sufficient level of security while enabling the effective embedding and extraction of information within

the steganographic framework. The model was thoroughly tested using various RGB image formats,

including both standard and non-standard ones, while accounting for different image capacities. The

experiments utilized a dataset of over 75 images, with the input size for testing set at 31,982 characters.

The performance analysis results of the proposed IRSA-PI-PVD model are outlined below. The sample

details of the input secret message are shown in Figure 2. The table 3 shows the sample input and sample

calculation for the encryption and PI_PVD calculation of the proposed model.
12
1.2. Assessment of Quality
Assessing the effectiveness of this model relies heavily on the quality of the stego image, which

represents the percentage of how well the human visual system can predict the presence of hidden

messages. Table 4 presents the experimental results of various evaluation parameters used to assess the

quality of this model.

The analysis from Table 4 demonstrates a direct relationship between MSE and RMSE, whereas PSNR

exhibits an inverse correlation with MSE. The average MSE and RMSE values are calculated as 2.6645

and 1.5553, respectively, indicating only a small deviation in pixel intensities compared to the cover

image. Furthermore, the average PSNR value of 41.5237 surpasses the basic limit of 30 dB, making it

difficult for intruders to detect steganography using the human visual system. Moreover, the average

SSIM value of 0.9629 indicates that the proposed model effectively preserves the structural and texture

information of the original image. Similarly, the average SC value of 0.9973 signifies that the stego image

maintains the original structural characteristics, and the concealed message introduces no noticeable

differences in pixel intensities compared to the cover image, which helps evade detection by intruders. In

summary, the average values of PSNR, SSIM, and SC quality parameters derived from the analysis affirm

that the proposed model successfully retains the structural characteristics and maintains the quality of the

cover image.

13
 Figure 2. Sample Input Secret Message

14
 Table 3. Sample Input, encryption, and PI-PVD calculation results of the proposed model
Random generated Encrypted
Secret Public Private R,G,B Color intensity values
Image Details values PI Secret
Message Key Pairs Key Pairs
p q r s Message Cover Image Stego Image
(176,37,92) (179,40,95)
(173,36,90) (173,33,88)
(175,36,91) (175,35,92)
pepper.jpg (761, (1566281, (177,38,93) (177,38,93)
"A" 31 53 29 37 Red 58944
(512x512) 1762939) 1762939) (179,38,91) (179,40,93)
(179,38,91) (179,36,90)
(180,37,91) (183,34,88)
(182,39,93) (182,41,95)
(105,103,52) (104,103,49)
(95,94,38) (95,94,40)
(126,125,69) (123,127,71)
baboon.jpeg (3953, (1330577, (136,136,76) (138,136,75)
"A" 53 41 43 17 Green 592923
(512x512) 1588463) 1588463) (128,128,66) (130,131,66)
(117,117,53) (115,117,53)
(101,102,36) (102,103,38)
(90,91,25) (89,91,23)
(105,103,52) (104,103,49)
(95,94,38) (95,94,38)
(126,125,69) (127,127,69)
baboon.jpeg (2527, (706207, (136,136,76) (136,136,76)
"Z" 67 47 7 53 Green 1003743
(512x512) 1168279) 1168279) (128,128,66) (127,131,65)
(117,117,53) (117,117,53)
(101,102,36) (100,103,38)
(90,91,25) (91,91,23)

15
 Table 4. Evaluations of various quality metrics of the proposed model

Input Size = 31982 Characters, Image Size = 512 x 512, Keys used: public key pair(3337,
446641), private key pair(361273, 446641)

Image Name Image Size PI MSE RMSE PSNR SC SSIM

pepper.jpg 512x512 Red 3.0383 1.7431 38.4781 0.9954 0.9522
pirate.jpg 512x512 Red 3.9582 1.9895 35.7621 0.9965 0.9526
babbon.jpeg 512x512 Green 3.3524 1.831 37.6236 0.9969 0.9623
camera.jpg 512x512 Red 3.2892 1.8136 37.7889 0.9966 0.9594
cat.jpg 512x512 Red 3.5073 1.8728 37.1286 0.9963 0.9732
moon.png 856x860 Green 0.6333 0.7958 52.0988 0.9992 0.9696
russel.jpeg 512x512 Blue 3.2114 1.792 37.997 0.9949 0.965
sun.jpeg 462x616 Red 4.5699 2.1377 34.9327 0.9943 0.9676
sun1.jpeg 1092x1092 Red 3.6041 1.8985 36.9948 0.9994 0.963
image.webp 600x900 Red 1.8176 1.3482 42.9407 0.9986 0.971
temple.jpg 1101x1500 Red 1.3673 1.1693 45.4137 0.9996 0.9598
treefog.jpg 427x640 Green 4.8608 2.2047 34.3966 0.9949 0.958
stones.jpg 427x640 Red 6.2536 2.5007 32.2082 0.9969 0.9532
parrot.jpg 853x1280 Red 1.5576 1.248 44.2817 0.9988 0.968
raindrop.jpg 512x743 Blue 1.7415 1.3197 43.3122 0.9953 0.9685
raining.jpg 512x1024 Blue 2.3725 1.5403 40.6267 0.9987 0.9702
circle.jpg 600x962 Red 1.3005 1.1404 45.8488 0.999 0.9584
eye.jpg 499x712 Red 1.7171 1.3104 43.4351 0.998 0.9621
lotus.webp 700x1200 Green 0.5946 0.7711 52.4736 0.9987 0.9598
jasmine.webp 1200x1600 Green 0.2707 0.5203 59.4804 0.9993 0.9742
dec_flrr.jpg 401x704 Green 2.9359 1.7135 38.7759 0.9962 0.9537
Average 634x836 - 2.6645 1.5553 41.5237 0.9973 0.9629

Table 5 presents a comparative analysis of the proposed model's performance against existing methods

based on key metrics such as Bits per Pixel (BPP), Peak Signal-to-Noise Ratio (PSNR), Structural

Similarity Index (SSIM), and embedding capacity. The data reveals that the proposed model achieves a

maximum embedding capacity of 2,527,068 bits, with an average PSNR of 38.48 dB, an SSIM value of

0.9522, and an embedding rate of 3.98 BPP, while embedding a 32 KB secret message in a standard 512 x

512 cover image.

(a) (b)

16
(c) (d)

(e) Histogram analysis of the green color channel (f) Histogram analysis of the green color channel
in the cover image in the stego image

(g) (h)

17
Figure 3. Histogram Analysis (a) Cover Image (b). Stego Image (c) Red channel intensity of the cover
image (d) Red channel intensity of the stego image (e) Green channel intensity of the cover image (f)
Green channel intensity of the stego image (g) Blue channel intensity in the cover image (h) Blue channel
intensity in the stego image

Table 5. Comparison of various quality metrics and embedding capacities of existing methods with the
proposed model
Image : pepper.jpg (size : 512 x 512)
Existing Models Embedding
BPP PSNR SSIM
Capacity
(Pradhan et. al. 2018) 2.08 43.13 0.9782 1171784
(Swain Gandharba 2018 b) 3.12 35.08 0.9581 2449935
(Swain, Gandharba 2018 a) 3.01 38.57 0.9561 2372858
(Swain, Gandharba 2018 October c) 3.14 38.74 0.9543 2470673
(Kalita et al. 2019) 2.31 37.43 0.9601 1816658
(Sonar R., Swain G 2021) 3.92 35.94 0.9541 3088513
(Monalisa Sahu et. al 2022) 1.98 36.53 0.973 1557135
(Supriadi Rustad et al. 2022) 1.541 35.6 0.9751 1211892
(Kaku Lishomwaa and Aaron Zimbab 2023) 1.663 40.23 0.9792 1307836
(Khadse, D.B., Swain, G 2023) 3.93 35.15 0.948 3091837
Proposed Model 3.98 38.48 0.9522 2527068

18
Figure 4 PDH – plots for the proposed IRSA-PI-PVD Model for the images (a) baboon.jpeg (b) pepper.jpg
(c) camera.jpg (d) cat.jpg (e) moon1.png (f) pirate.jpg
19
 This analysis highlights that the proposed method offers a significantly higher BPP compared to

existing approaches while maintaining a PSNR of 38.48 dB and an SSIM value slightly below 1. Despite

being marginally lower than some existing methods, the SSIM remains close to 1, indicating that the

proposed technique preserves the structural integrity of the image with minimal distortion. The PSNR

value of the proposed model, which exceeds the generally accepted threshold of 30 dB, suggests that the

alterations introduced by the embedding process are imperceptible to the human visual system, even with

an enhanced embedding capacity. This ensures that the steganographic content remains undetectable to the

naked eye while providing substantial improvements in terms of a 3.98 BPP.

Furthermore, the proposed model integrates the IRSA cryptography technique, enhancing the

security of the embedded messages during transmission. By employing this cryptographic method, the

confidentiality of the secret data is significantly strengthened, requiring intruders to invest considerable

time and effort in cryptanalysis, even if they are able to detect the embedding technique. The results

clearly demonstrate that the proposed model achieves notable improvements in both embedding capacity

and the quality of the stego image, offering superior performance in terms of BPP, PSNR, and SSIM

compared to existing techniques.

1.3. Assessment of Security

1.3.1. Visual Analysis

The comparison between the stego image and the cover image relies on the human visual system's

quality, determining its ability to detect the hidden message in the stego image. The proposed model

achieves an average embedding capacity of 101,356 bits and an average PSNR value of 37.3563dB for

different images with a standard size of 512x512. The small changes in intensities make it difficult for the

20
human visual system to detect alterations in the stego image. As a result, intruders cannot perceive any

changes in the cover image.

1.3.2. Histogram Analysis

Using histogram representation, the anomalies in the stego image concerning the cover image are

identified based on the pixel intensity values of the red, green, and blue channels. The 2D graph is plotted,

with the X-axis representing the order of the pixel location in the image and the Y-axis representing the

intensity values. Figure 3 displays the histogram representation for the image pepper.jpg. The analysis

from Table indicates that the average values of MSE and RMSE are minimal. As a result, intruders are

unable to detect the small deviations in the intensities of the various channels, thereby making it difficult

for them to predict the hidden messages in the secret message through visual analysis.

1.3.3. Peak Difference Histogram (PDH) Analysis

The PDH analysis is performed to determine the difference in pixel intensities between consecutive

pixels in both the cover image and the stego image. The values of pixel intensity differences varies from -

255 to 255, including 0. A histogram is plotted based on the frequency count of each difference value the

X-axis representing the pixel difference and the Y-axis representing the frequency count. 4, represents the

PDH analysis of the various images. The PDH curves between the cover and stego images display a

smooth curve, indicating that the steganography remains undetected by intruders. This suggests that the

pixel intensity differences do not reveal any discernible patterns that would lead to the detection of hidden

messages in the stego image.

1.4. Execution Time Analysis

The proposed model enhances the traditional RSA algorithm by utilizing four large prime numbers p,

q, r, and s instead of the conventional two, significantly increasing the complexity of the factorization

problem. These primes are crucial for determining the security strength of the RSA algorithm. For optimal
21
security, the prime numbers must be distinct, randomly generated, and sufficiently large, typically

between 2048 and 4096 bits, as recommended for modern cryptographic applications. By expanding the

keyspace and introducing four primes, the model not only strengthens resistance to brute force attacks,

shared factor attacks, and cryptanalytic vulnerabilities, but also reduces the risk of common factorization

attacks. Moreover, this approach improves quantum resistance, effectively mitigating the threat posed by

future quantum-based decryption methods, thus enhancing the overall security and reliability of the

encryption process.

In the context of time analysis, comparing RSA with IRSA is essential for evaluating the efficiency and

performance improvements that IRSA offers over the traditional RSA algorithm. This comparison helps

identify how specific enhancements in the IRSA methodology affect encryption and decryption times,

which are critical indicators of effective security in real-time applications. Furthermore, when examining

the proposed embedding technique, it is important to note that the time required for embedding and

extracting secret message bits using IRSA remains relatively consistent when compared to standard PVD

or any newer PVD models. This finding suggests that implementing the IRSA model does not

significantly increase the computational overhead associated with these embedding and extraction

processes. Consequently, it facilitates efficient data hiding without compromising the speed of operations.

By maintaining comparable time consumption rates across various PVD models, the proposed IRSA

embedding technique illustrates its capability to integrate seamlessly into existing steganographic

frameworks, ensuring that both security and performance are optimized within the realm of image

steganography.

Table 6 presents the execution times of both the standard RSA and IRSA algorithms across different

file sizes. The data depicted in Figure 5 indicates that the IRSA algorithm requires more time for both

encryption and decryption, with embedding and extraction times remaining relatively consistent. This

discrepancy in time allocation suggests that the proposed encryption algorithm offers enhanced security in

22
comparison to the conventional RSA algorithm. Consequently, unauthorized person encounters a

heightened level of difficulty in their endeavours to uncover plaintext content within encrypted messages.

Table 6. Execution Time Comparison of the Standard RSA Algorithm and the Improved RSA Algorithm
for Different Input Sizes
Total time taken Total time taken Total time taken Total time taken
Total No. for Encryption for Embedding for Extraction for Decryption
File
of bits (in ms) (in ms) (in ms) (in ms)
Name
embedded Standar Standar Standard IRS Standard
IRSA IRSA IRSA
d RSA d RSA RSA A RSA
test11.tx
73132 62.5 687.2 453.1 468.7 187.9 187.5 1510.6 5857.9
t
test12.tx
132544 140.6 1280 828 845.7 373.7 374.9 2532.6 10503
t
test14.tx
296030 296.8 2812 1827.8 1917 578 593.9 8690.9 22998
t
test13.tx
464602 456.1 4393 2942.9 2981 1778.7 1803 17428.1 36943
t
test17.tx
526476 505 4859 3265 3348 999.9 1062 21936.4 40551
t
test16.tx
695234 684.6 6405 4329.8 4401 1342.1 1375 32436 53562
t
test15.tx
894688 859.1 8328 5530.2 5577 1746.9 1765 47153.8 68310
t
test18.tx
1026202 1015.4 9010 6395.9 6400 1960.1 5382 58723.5 82032
t

CONCLUSION
The experimental results underscore the model's ability to enhance the embedding capacity to 1007996

bits and elevate the average PSNR value to 37.1286 dB for standard 512x512 images. In terms of security

considerations, the model outperforms various vulnerability attacks, providing a heightened level of

security. Additionally, the approach ensures extended cryptanalysis duration to decipher the original

message from the cipher text. Through the integration of cryptographic and steganographic techniques, the

model establishes a robust four-tier security framework for safeguarding confidential messages within the

23
spatial domain of image steganography. It is recommended to subject the model to diverse cryptographic

and steganographic techniques to comprehensively evaluate its performance.

Figure 5 Execution time analysis between standard RSA and Improved RSA (a) Encryption time
analysis (b) Embedding time analysis (c) Extraction time analysis (d) Decryption time analysis

REFERENCES

24
Abdulhammed, Omar Younis. 2023. “A Robust Image Steganography Based on a Novel Technique

by Using Improved DNA and Modified Chaotic Approach.” The Journal of Supercomputing,

Springer Science and Business Media LLC, June 2023. doi:10.1007/s11227-023-05459-x.

Alexander G., Chefranov, and Gurcu Oz. 2002. “Adaptive to Pixel Value and Pixel Value

Difference Irreversible Spatial Data Hiding Method Using Modified LSB for Grayscale Images.”

Journal of Information Security and Applications, 70 (1): 103314, Nov. 2022,

doi:10.1016/j.jisa.2022.103314.

Amalarethinam, I. George, and H. M. Leena. 2017. “Enhanced RSA Algorithm With Varying Key

Sizes for Data Security in Cloud.” 2017 World Congress on Computing and Communication

Technologies (WCCCT), Feb. 2017, doi:10.1109/wccct.2016.50.

Dharwadkar, N.V., Mahmud, M., Lonikar, A.A., Brown, D.J. 2023. “A Medical Image

Steganography Scheme With High Embedding Capacity to Solve Falling-Off Boundary Problem

Using Pixel Value Difference Method.” Communications in Computer and Information Science,

Springer Nature Singapore, 320–32. doi:10.1007/978-981-99-1648-1_27.

Gandharba Swain 2019,"Two new steganography techniques based on quotient value differencing

with addition-subtraction logic and PVD with modulus function" Optik, 180: 807-823 Feb. 2019,

https://doi.org/10.1016/j.ijleo.2018.11.015.

H.-C. Wu, N.-I. Wu, C.-S. Tsai, and M.-S. Hwang. 2005. “Image Steganographic scheme based on

pixel-value differencing and LSB replacement methods,” IEE Proceedings - Vision, Image, and

Signal Processing, 152(5), 2005. doi:10.1049/ip-vis:20059022

Kaku Lishomwaa and Aaron Zimbab. 2023. “A Privacy-Preserving Scheme for Medical Diagnosis

Records Based on Encrypted Image Steganography.” Zambia ICT Journal, 7(1): 23–28, Mar.

2023, doi:10.33260/zictjournal.v7i1.151.

25
Kalita, Manashee, Themrichon Tuithung, and Swanirbhar Majumder. 2019. “An Adaptive Color

Image Steganography Method Using Adjacent Pixel Value Differencing and LSB Substitution

Technique.” Cryptologia, 43(5) :414–37, Apr. 2019, doi:10.1080/01611194.2019.1579122.

Khadse, D.B., Swain, G 2023, “Data Hiding and Integrity Verification based on Quotient Value

Differencing and Merkle Tree” Arabian Journal for Science and Engineering, 48: pp.1793–1805

July 2023. doi: https://doi.org/10.1007/s13369-022-06961-9

Monalisa Sahu, Neelamadhab Padhy and Sasanko Sekhar Gantayat. 2022. “Multi-directional PVD

Steganography Avoiding PDH and Boundary Issue.” Journal of King Saud University -

Computer and Information Sciences, 34(10) : 8838–51, Nov. 2022,

doi:10.1016/j.jksuci.2021.10.007.

Nasution, Nadia, Efendi, Syahril and Nasution, Sawaluddin. 2020. “Analysis of RSA Variants in

Securing Message.” IOP Conference Series: Materials Science and Engineering, 725(1): 012131,

Jan. 2020, doi:10.1088/1757-899x/725/1/012131.

Pradhan, Anita, Sekhar, K. Raja, Swain, Gandharba. 2018. “Digital Image Steganography Using

LSB Substitution, PVD, and EMD, Mathematical Problems in Engineering.” Hindawi

Mathematical Problems in Engineering 2018, 1804953, 11 pages, September 2018, doi:

https://doi.org/10.1155/2018/1804953

Setyono, D. R. I. M. Setiadi and Muljono 2017. “StegoCrypt method using wavelet transform and

one-time pad for secret image delivery." 2017 4th International Conference on Information

Technology, Computer, and Electrical Engineering (ICITACEE), Semarang, Indonesia, 203 -

207, 2017, doi: https://10.1109/ICITACEE.2017.8257703

Sonar, R., Swain, G. 2021, Steganography based on quotient value differencing and pixel value

correlation. CAAI Transactions on Intelligence Technology, 6(4): pp.504 - 519 April 2021.

https://doi.org/10.1049/cit2.12050

26
Supriadi Rustad, De Rosal Ignatius Moses Setiadi. 2022. “Inverted LSB Image Steganography

Using Adaptive Pattern to Improve Imperceptibility.” Journal of King Saud University -

Computer and Information Sciences, 34(6) : 3559–68, June 2022,

doi:10.1016/j.jksuci.2020.12.017.

Supriadi Rustad, Ignatius Moses Setiadi De Rosal, Pulung Nurtantio Andono, Abdul Syukur, and

Purwanto. 2022. “Optimization of Cross Diagonal Pixel Value Differencing and Modulus

Function Steganography Using Edge Area Block Patterns.” Cybernetics and Information

Technologies, 22(2): 145–159, June 2022, doi: https://doi.org/10.2478/cait-2022-0022.

Swain, Gandharba. 2015. “Adaptive Pixel Value Differencing Steganography Using Both Vertical

and Horizontal Edges.” Multimedia Tools and Applications, 75 (21): 13541–56, Oct. 2015,

doi:10.1007/s11042-015-2937-2.

Swain, Gandharba 2018, “High Capacity Image Steganography Using Modified LSB Substitution

and PVD against Pixel Difference Histogram Analysis.” Security and Communication Networks,

Article ID 1505896, 14 pages, September 2018. doi: https://doi.org/10.1155/2018/1505896

Swain, G Gandharba 2018, “Adaptive and Non-adaptive PVD Steganography Using Overlapped

Pixel Blocks” Arabian Journal for Science and Engineering”, 43(1), pp.7549–7562 Feb. 2018.

https://doi.org/10.1007/s13369-018-3163-9

Swain, Gandharba 2018, “A data hiding technique by mixing MFPVD and LSB substitution in a

pixel” Information technology and control, 47(4):714-727 Cotober 2018. doi:

https:10.5755/j01.itc.47.4.19593.

Thanikaiselvan V., Selvaraj, Subashanthini and Amirtharajan R. 2014. “PVD Based Steganography

on Scrambled RGB Cover Images With Pixel Indicator.” Journal of Artificial Intelligence, 7(2):

54–68, Mar. 2014, doi:10.3923/jai.2014.54.68.

27
Tseng, Hsien-Wen, and Hui-Shih Leng. 2013. “A Steganographic Method Based on Pixel-Value

Differencing and the Perfect Square Number.” Journal of Applied Mathematics, 2013 : 1–8,

2013, doi:10.1155/2013/189706.

Wu, Da-Chun, Wen-Hsiang Tsai. 2003. “A Steganographic Method for Images by Pixel-value

Differencing.” Pattern Recognition Letters, 24 (9–10) :1613–26, June 2003, doi:10.1016/s0167-

8655(02)00402-6.

Zhang, Xinpeng, and Shuozhong Wang. 2004. “Vulnerability of Pixel-value Differencing

Steganography to Histogram Analysis and Modification for Enhanced Security.” Pattern

Recognition Letters, 25(3): 331–39, Feb. 2004, doi:10.1016/j.patrec.2003.10.014.

28