Spatial Domain-Based Robust Watermarking Framework for Cultural Images

Heritage multimedia, which include photographs, customs, knowledge, arts, rituals, audio, cultural information, and music, are valuable artifacts of any region. The most important attribute of heritage media is the transmission of important features of past generations, which reflect their way of living, innovative attitude, and diversity in archaeological and historical perspectives. However, the proliferation of the Internet has made such data exchange more challenging than ever, allowing unauthorized users to easily access such information. Under such circumstances, securing cultural heritage (CH) media is essential. In that regard, herein, we present a spatial domain-based blind and robust watermarking scheme for the ownership verification of colored CH images; this scheme uses DC coefficient modification. In this scheme, the “Y” element of the YCbCr space is used for inserting a watermark. The “Y” element of a host image is divided into non-overlapping blocks with sizes of <inline-formula> <tex-math notation="LaTeX">$8\times8$ </tex-math></inline-formula>. Each <inline-formula> <tex-math notation="LaTeX">$8\times $ </tex-math></inline-formula> 8 block is then divided into two <inline-formula> <tex-math notation="LaTeX">$4\times $ </tex-math></inline-formula> 8 subblocks. Instead of calculating the DC coefficients using the discrete cosine transform, we independently calculate the DC coefficient of every <inline-formula> <tex-math notation="LaTeX">$4\times $ </tex-math></inline-formula> 8 subblock in the spatial domain. We test our method based on standard test images obtained from the USC-SIPI dataset and a self-created dataset of cultural images. Our scheme demonstrates improved robustness and lower computational complexity than frequency-domain-based techniques. The average peak signal-to-noise ratio of the proposed technique for test images is 40.0830 dB, and the structural similarity index matrix value is closer to one under no attack, ensuring the imperceptibility of the technique. Further, we prove the resilience of the proposed algorithm by comparing it with various state-of-the-art techniques.


I. INTRODUCTION
The term ''heritage'' refers to the cultures, qualities, and traditions in a region/country that have prevailed over generations and are of great significance to the country. Cultural The associate editor coordinating the review of this manuscript and approving it for publication was Zijian Zhang . heritage (CH) is a way of livelihood that mankind has inherited from prior generations and is circumvented to the following generations. CH includes natural heritage (culturally remarkable biodiversity and landscapes), intangible culture (festivals, knowledge, oral traditions, expressions, rituals, and languages), and tangible culture (traditional clothing, artifacts, books, and monuments) [1]. It reinforces the recognition of community culture, thereby enhancing the social economy, and must be leveraged by cultural industries under the protection of intellectual property rights (IPR). Although CH symbolizes an essential asset of a specific society, region, or nation, a justifiable means of its protection is digitization, as proposed by UNESCO in the Convention for Safeguarding the Cultural Heritage in 2003. In addition to ensuring its preservation for future generations, digitization certifies global passage to the various cultures of global heritage and preserves its priceless assets from degradation. However, owing to the rapid advancement in the Internet and the rapid development of multimedia-oriented systems in a variety of fields, access to CH media has become easier. In particular, the development of virtual galleries has allowed people to access information and immerse themselves in art without physical barriers. People of all ages can participate in exhibitions remotely, making them more responsive to learning. However, with these benefits, the latest technologies could also adversely influence digital information by permitting unauthentic data processing, fake information, and illegitimate copying of valuable artifacts, such as statues, buildings, paintings, and other heritage data from galleries. Notably, when an artwork is shared digitally, it is shared in the most appealing manner to attract others. By exploiting this intent, an unauthorized user can copy the original content and make an illegal sale. Therefore, the need of the hour is to use data protection techniques to secure digital multimedia content from unauthorized distribution and illegal copying to provide ultimate protection to artifacts [2]. In that regard, among the numerous methods adopted to protect CH data, digital watermarking has attracted considerable attention in the current scenario owing to its potential in certifying data authentication and validating ownership rights. Using this method, unrevealed data (called watermarks) are hidden inside digital data without threatening the confidentiality of the original data [3]. Digital watermarking methods can be typically categorized into two types: robust and fragile. Among these, fragile watermarking can be used to achieve integrity protection [4]; herein, the watermark cannot ideally be extracted if the watermarked cover image has been tampered with. In contrast, the basic motive behind robust watermarking is ownership protection. Here, even if the watermarked host is modified, the watermark can be extracted. Because the field of robust watermarking is primarily used with the aim of ensuring ownership security, it is interpreted to terminate attacks that attempt to demolish a watermark without appreciably derogating the perceptual representation of the watermarked image. In the process of digital watermarking, robustness is one of the major parameters to be considered, that is, the watermark extracted should be robust enough to guarantee the ownership of the cover image, although the watermarked image may be disclosed to various signalprocessing operations. Although various watermarking techniques for improving the resilience of a watermarked picture have been proposed in the literature, designing a robust system that can achieve spectacular robustness and support a better visual representation of watermarked images remains a concern in the field of watermarking.
In particular, the transmission and processing of colored images play an indispensable role in the present informationacquainted civilization. Therefore, greater consideration must be given to color images than grayscale images. Notably, various color models can be utilized for image watermarking; however, the two basic color models used are RGB and YCbCr. The RGB color model is preferred for displaying colors observed in the natural world. On the contrary, the YCbCr color model separates visual information into three constituents: one luminance (Y) and two chrominance constituents (Cb and Cr). Previous studies have indicated that the YCbCr color model demonstrates improved robustness against various signal processing and geometric attacks than the RGB color space model [1]; therefore, we chose to explore it in this study.
Thus, in this paper, a robust, blind, and computationally effective authentication-based watermarking technique is proposed; the technique computes the DC coefficient of a block without using the discrete cosine transform (DCT); this is because the primary motive of this study is to protect the ownership rights of CH images. In this technique, the ''Y'' element of a host image is divided into non-overlapping blocks with sizes of 8 × 8; following this, every 8 × 8 block is divided into two 4 × 8 subblocks. Instead of calculating the DC coefficient using the DCT, we independently calculate the DC coefficient of every 4 × 8 subblock in the spatial domain. Following this, watermark bits are inserted in the spatial domain by altering the DC coefficients of several subblocks. The proposed approach demonstrates improved robustness than frequency-domain-based techniques. In addition, the scheme is computationally less complex, because embedding is performed in the spatial domain. The YCbCr color model is used, rather than the RGB color model, to embed the watermark in the spatial domain. Further, the Y channel is utilized for watermark embedding owing to its high robustness.
The remainder of this paper is organized as follows. A review of relevant literature is presented in Section II. Section III presents the limitations of previous studies and the objectives of the present study. The mathematical preliminary framework is established in Section IV. The proposed scheme is described in Section V. Section VI presents the experimental results, and a discussion of the results is presented in Section VII. Finally, the conclusion is presented in Section VIII.

II. RELATED RESEARCH
Numerous watermarking schemes for the protection of concealed information, content authentication, and IPR protection have been proposed in numerous studies, either in the spatial or transform domain [3], [4], [5]. This section presents a summary of such previously proposed watermarking strategies. In [6], a watermarking system based on the DCT was proposed by utilizing the psycho-visual threshold standard. The bits of the watermark were placed in the cover VOLUME 10, 2022 picture by changing the correlation coefficients drawn using a predefined rule. The approach offered better resilience; however, the cover picture used was grayscale. In [7], the authors suggested a binary-tree-based quantization wavelet domain watermarking scheme. To produce a watermarked image, the technique constructed a hierarchical watermarked image/video code stream that could be truncated at any distortion-robust atom. In [8], a blind hybrid watermarking technique using the discrete Fourier transform (DFT) and discrete wavelet transform (DWT) was reported. The algorithm offered better robustness against common signal-processing operations; however, it has not been tested against any hybrid attacks. Khafaji et al. [9] proposed a robust watermarking algorithm by using selected graph Fourier coefficients to embed a watermark. The proposed scheme demonstrated improved robustness against various attacks by establishing a relationship between watermark extraction. In [10], a watermarking technique utilizing the DCT and a reiteration code was presented. Although simultaneous operations are yet to be investigated, this scheme offered resistance against common geometrical operations. In [11], the authors proposed a DWT-based dual watermarking strategy, wherein the YCbCr color space was used for embedding a robust watermark, and a fragile watermark was inserted in the RGB color space utilizing an improved form of the least-significant-bit (LSB) substitution procedure. The algorithms resulted in greater computational complexity. In [12], the authors introduced a two-dimensional (2D) DCT-based watermarking framework that incorporated a pseudorandom sequence to insert the watermark into the middle-frequency coefficients of a colored picture. Although this framework demonstrated better performance against common signal-processing attacks, no tests have been performed for hybrid attacks. In [13], a watermarking technique relying on the DFT was introduced, wherein different types of Fourier transforms (fractional FT, DFT, and quaternion FT) were used, and the parity of outcome values were utilized for inserting the watermark. In [14], a dual domain-based watermarking method was proposed, and herein, the YCbCr color model was used for embedding dual watermarks, that is, a robust watermark was embedded in the Y component using the DCT, and a fragile watermark was inserted in the Cb component. Although this scheme demonstrated good robustness, it resulted in a higher computational complexity. In [15], the author proposed a DCT-based watermarking strategy, wherein the grayscale and two color spaces (YCbCr and RGB) were used to insert the watermark using the middle-frequency values of the cover picture. In addition, Arnold's transform and chaotic encryption techniques were used to improve the watermark security. A spatial domain-based watermarking algorithm was introduced in [16], wherein the cover picture used was a grayscale image, and watermarks were directly embedded into the DC coefficients. However, this algorithm is yet to be investigated for combined attacks. In [17], a robust watermarking algorithm utilizing the DCT was proposed, wherein the B channel of the RGB color model was used for watermark embedding based on the quantified DCT coefficient selection method. Although this scheme resulted in a better peak signal-to-noise ratio (PSNR), it lacked robustness. In [18], the author proposed a robust reversible watermarking algorithm for encrypted images. The watermark was embedded using a prediction error expansion procedure based on a protected multiparty computation method. However, the framework has not been investigated for any combined attacks. In [19], a spatial domain watermarking algorithm was proposed for the ownership security of color images using a DC-coefficientbased quantization watermarking procedure. Although the scheme offered less computational complexity, its robustness was not up to mark. In [20], the authors introduced quaternion singular value decomposition (QSVD) and quaternion wavelet transform (QWT)-based watermarking algorithms using the YCbCr color space. A 2D Chebyshev-logistic map was also incorporated to encrypt watermark to enhance the security. However, the scheme was found to be robust against only common signal-processing attacks. In [21], the authors presented a robust watermarking strategy utilizing a combination of two transforms, that is, a redundant DWT and non-subsampled contourlet transform, to insert a watermark. The scheme achieved better imperceptibility; however, it was not resilient to various signal-processing and geometrical attacks.
So far, a few studies on the protection of CH multimedia have been reported in the literature. In [22], the authors used an application-based watermarking strategy, wherein a pseudorandom sequence was used in frequency domain-based DFT coefficients to insert a watermark in the cover image. However, no signal-processing attacks were performed on the scheme to test its robustness. An integrated software (LCI)-based watermarking algorithm was introduced in [23]; this algorithm was designed to provide protection to artifacts using a robust watermarking scheme. Here, DFT-based midfrequency coefficients were used to embed a watermark in the frequency domain of the cover picture. However, this technique has not been investigated for signal-processing attacks. In [24], the authors introduced a DCT-DWT-based dual domain watermarking method for CH data protection using semi-fragile watermarking techniques. The YIQ color space (where Y stands for luminance, and I and Q denote inphase and quadrature components, respectively) was used to investigate different applications, such as data authentication, image compression, error correction, and copyright protection. However, this technique was tested only for singular attacks. In [25], the author adopted the difference expansion of Tian's algorithm as the primary technique for watermarking in a reversible format and also incorporated channel and lifting encoding to secure CH pictures. A lifting-based twolevel DWT in the transform domain was incorporated as a transformation tool to embed a watermark. However, the proposed technique is unsuitable for real-time applications owing to its high mean running time of 48.55 s. In [26], the author proposed a robust watermarking algorithm that used the principal constituents of multichannel images, instead of watermarking through their channels. In the principal constituent technique, the watermark was placed in the strongest element of each channel in the image for protection. The scheme was computationally efficient; however, it has not been tested for robustness.

III. SHORTCOMINGS OF PREVIOUS STUDIES AND OBJECTIVES OF THE PROPOSED SCHEME
A few shortcomings of previous studies reported in the literature are listed below: • Several watermarking systems presented in the available literature perform better against common signal-processing attacks but exhibit less resilience against different geometric attacks.
• The capacity and imperceptibility of most of the techniques are not up to the mark.
• Limited techniques reported in the literature have been investigated for their computational complexity.
• The majority of techniques introduced for CH image authentication and IPR protection are based only on grayscale images.
• Generally, the schemes reported in previous studies for robust watermarking have been examined explicitly against singular attacks, and very little attention has been paid to the analysis of simultaneous attacks.
Therefore, considering the shortcomings of previous methods, as listed in Table 1, we developed a time-efficient, blind, and robust authentication-based watermarking algorithm in this study; this method demonstrates resilience for copyright protection of CH images while simultaneously addressing all the aforementioned limitations. The contributions of this study can be summarized as follows: • The system was made robust by inserting watermark bits in the spatial domain instead of embedding them in the transform domain.
• The watermark was embedded in the ''Y'' element of the YCbCr space, which offered better resilience, and the scheme could be used for applications involving visual sensors.
• The proposed scheme offered better resistance against various simultaneous and singular attacks compared with various other state-of-the-art schemes.
• One of the most important characteristics of this scheme is its low computational complexity, which makes it suitable for real-time applications.

IV. DCT
The DCT is widely used to convert a signal from the spatial domain to the transform domain. For digital watermarking applications, DCT-based transformation is typically used because it employs the standard compression technique (JPEG).
For an image with a size of M × N , the 2D DCT can be computed as follows: where v and u denote the vertical and horizontal frequency components, respectively. The inverse DCT transforms a signal from the frequency domain back into the spatial domain using (3), as follows: From (1), the DC coefficient can be calculated as follows: Therefore, from (4), the DC coefficient T (0, 0) can be calculated by directly considering the mean summation of the overall pixels of s(i, j) in the spatial domain. Thus, it can be observed that adding a watermark directly to the DC coefficient does not afflict any losses after the application of the inverse DCT [16]. In our study, we mimicked the transform domain behavior by implementing the algorithm in the spatial domain. Notably, the proposed technique required less computational time, similar to techniques based on the spatial domain, and simultaneously provided robustness, similar to those based on the transform domain. This is because we did not use the actual DCT for the transformation of the cover image; instead, we used the fundamental concept directing that the DC coefficient of a transformed image can be computed by determining the mean of its intensities.

V. PROPOSED FRAMEWORK
As stated, herein, a blind and robust authentication-based watermarking technique operating in the spatial domain is proposed for CH images. The watermarking system is divided into two steps: embedding and extraction processes.
A. EMBEDDING PROCESS Fig. 1 presents a block diagram of the proposed watermark embedding process. The cover image is first converted into the YCbCr space from the RGB space, and the luminance element ''Y'' is selected for hiding the watermark. In the ''Y'' element, watermark insertion is executed according to the following steps: Step 1: Split a 512 × 512 ''Y'' element into nonoverlapping blocks with sizes of 8 × 8, where ''A'' denotes a random 8 × 8 block. Step 2: Split the block ''A'' again into two subblocks, each with a size of 4 × 8, consequently dividing the odd and even location pixels, as depicted in Fig. 2. The pixel subblocks of odd and even locations are represented as A odd and A even , respectively.
Step 3: Calculate the DC coefficients of A odd and A even using (4). The obtained DC coefficients are represented by D odd and D even , as expressed in (5) and (6), respectively.
Here, (5) and (6) indicate that the DC coefficient of a block can be directly computed based on the mean summation of all pixels in that block, instead of calculating the DCT of a block. A watermark is inserted by changing the DC coefficients of A odd and A even so that D odd becomes greater than D even for watermark bit ''1'' and vice versa for watermark bit ''0.'' Step 4: Insert watermark bit ''0'' or ''1.'' The watermark bit ''1'' is embedded using the following expression: If D odd < D even , then D odd = D even D even = D odd

End
To increase the resilience of the system, the difference between the two DC coefficients is maintained greater than a predefined embedding factor µ, as follows: If D odd − D even < µ then

End
To increase the resilience of the system, the difference between the two DC coefficients is maintained greater than a predefined embedding factor µ, as follows: If D even − D odd < µ then End End of embedding of bit ''0'' Step 5: Determine the magnitude of change resulting from the pixel values of A odd and A even , as follows: Step 6: Obtain the changed subblocks as follows: A * even = A even + ∅ even (10) Step 7: Combine A * odd and A * even to obtain the watermarked blockA * .
Step 8: Repeat Steps 2-7 until all bits of the watermark are inserted in the ''Y'' element blocks, resulting in the watermarked luminance element. The value of µ used in this experiment is 20. Further, we also conducted an experiment using various values of µ. Increasing the value of µ above 20 resulted in better robustness but degraded the perceptual quality of the image; by contrast, decreasing µ below 20 improved the perceptual quality of the image but degraded the robustness of the algorithm. Thus, we selected an optimum value of µ to retain the robustness and visual quality of the image. The final watermarked image was obtained after converting the watermarked ''Y'' element to the RGB space from the YCbCr space.

B. EXTRACTION PROCESS
The extraction process of the digital watermark is similar to the watermark embedding process. Fig. 3 presents a block diagram of the extraction process. The watermarked picture was converted to the YCbCr space from the RGB space, where the luminance component was forwarded for watermark extraction. The steps involved in the extraction of the watermark from the watermarked luminance component can be outlined as follows: Step 1: Divide the 512 × 512 watermarked ''Y'' element into non-overlapping blocks with sizes of 8 × 8, where ''A'' denotes an arbitrary 8 × 8 block.
Step 2: Split the block ''A'' into two subblocks with sizes of 4 × 8, dividing the odd and even location pixels, as indicated in Fig. 3. The pixel subblocks of odd and even locations are represented by A odd and A even , respectively.
Step 3: Calculate the DC coefficients of A odd and A even using (5) and (6), respectively. Suppose that the DC coefficients of A odd and A even are represented using D odd and D even , respectively.  Step 4: Follow the watermark extraction steps to obtain the watermark bits as follows: If D odd < D even ,then Watermark bit = 0 Otherwise Watermark bit = 1

End
Step 5: Repeat Steps 2-4 until the bits of the watermark are obtained from all the blocks of ''Y'' and produce the extracted watermark.

VI. EXPERIMENTAL RESULTS
We conducted our experimental analysis on the following test images: ''Airplane'' and ''Lena'' with a size of 512 × 512 procured from the USC-SIPI database and 176 CH  images with a size of 512 × 512 extracted from our selfcreated database; a few of them are labeled as Image ''A'' to Image ''F.'' In addition, we used watermarks with various sizes for the experimental analysis. Fig. 4 presents a few test images and watermarks used in the experiment. Various evaluation quality parameters, such as the normalized correlation coefficient (NCC), similarity index matrix (SSIM), bit error rate (BER), and PSNR, were utilized for the proposed framework analysis [15].

A. IMPERCEPTIBILITY ANALYSIS
Note that any watermarking algorithm must be imperceptible; that is, after embedding the watermark, the visual quality of the cover image should not change, and the watermarked image should not appear degraded compared to the actual cover image. The watermarked images without attack  subjugation, along with their extracted watermarks, are presented in Fig. 5, where the NCC is equal to one. The attacked and extracted images presented in Fig. 5 reveal that the proposed embedding algorithm demonstrates good visual quality. Table 2 lists the SSIM and PSNR values of the standard test and CH images obtained under a no attack scenario.
The achieved PSNR for watermarked images was approximately 40 dB, indicating the ability of this technique in providing a better perceptual quality of watermarked images.

B. ROBUSTNESS ANALYSIS
Robustness is a basic parameter that indicates the resistance of an algorithm to different geometric and signal-processing operations. The robustness is measured using parameters such as the NCC and BER. In our experiments, the test and CH images were subjected to various singular attacks (including ''salt and pepper,'' ''JPEG compression,'' ''filtering,'' ''sharpening,'' ''speckle noise,'' etc.) and hybrid attacks, and the objective analysis is presented in Tables 3 and  4. Fig. 6 illustrates the images with the applied watermarks.  The results of the performance analysis of the algorithm after various simultaneous attacks in terms of the NCC are presented in Fig. 7. In addition, the efficiency of the algorithm is compared with that of multiple state-of-the-art schemes, and the corresponding results are presented in Table 5 (a), Table 5 (b), Fig. 8, Fig. 9, and Fig. 10.

C. PAYLOAD
Notably, payload defines the number of watermark bits that can be inserted into a host image. In the proposed scheme, the host image used a 24-bit color image (RGB image), which was then converted into the YCbCr color space, where one watermark bit was embedded in each 8 × 8 block of the Y component of the host image. Therefore, the payload varied with the size of the host image. Table 7 summarizes the payload of the proposed scheme for host images with different sizes. VOLUME 10, 2022

D. COMPLEXITY
An Intel (R) core (TM) i7-7700 central processing unit operating at 3.60 GHz was used for objective assessment of the proposed framework on MATLAB 2017a with a Windows operating system, and the obtained results are listed in Table 8. In this table, the average time (in seconds) is provided for a few test images. A timing comparison of the proposed scheme with that proposed in [14] is presented in Fig. 11. The scheme proposed in [14] was first tested under the same experimental conditions as the proposed algorithm. After maintaining the same experimental conditions, we compared the computational time of the proposed scheme with that of the state-of-the-art technique.

VII. DISCUSSION
As stated, the proposed technique was tested for its visual efficiency and robustness using evaluation parameters such as the PSNR, SSIM, and NCC. The mean PSNR of the proposed scheme under a no attack scenario was greater than 40 dB, and the SSIM values were closer to one, as listed in Table 2, which ensured the imperceptibility of the proposed framework. Fig. 5 and Table 2 present the perceptual quality of the proposed scheme. The proposed technique could insert a total of 4096 bits of a watermark in a host image with a size of 512 × 512 by inserting one bit in every 8 × 8 block, which is further illustrated in Table 7. The proposed scheme was    Tables 3 to 5 are either closer to one or equal to one, which indicates that the obtained watermarks can still be recognized after being subjected to various singular and simultaneous attacks. The proposed scheme (Tables 5 (a) and 5(b)) was then compared with various well-known state-of-the-art schemes, revealing that the proposed technique offered better resilience than the available watermarking schemes. In addition, the proposed framework

FIGURE 11.
Comparison of the timing parameters of the proposed framework with those of the technique proposed in [14].
was evaluated for different watermarks with varying sizes, and the results are presented in Table 6; these results also indicate the resilience of the scheme for different-sized watermarks. The results of the timing analysis of the framework are presented in Table 8. Furthermore, the results of the timing comparison of the proposed algorithm with that proposed in [14] (Fig. 11) also prove the capability of the technique for use in real-time applications.

VIII. CONCLUSION
CH images are crucial assets of any region. In the ongoing scenario of Internet dominance, any unauthorized user can effortlessly access valuable data and alter their original ownership. Thus, the security and protection of CH data in such scenarios are significant challenges for researchers worldwide. Therefore, it is the need of the hour to ensure copyright protection of CH images. To that end, we developed a blind and robust pixel-domain-based watermarking scheme that offered better imperceptibility and robustness. The performance of our scheme was investigated for different image processing operations, such as salt and pepper noise, low-pass filtering, sharpening, and hybrid attacks. The experimental results indicate that in addition to offering resilience against singular attacks, the proposed algorithm also offers robustness against simultaneous attacks. The comparison analysis reveals that our proposed technique performs better in terms of the robustness and imperceptibility compared to various state-of-the-art techniques, while demonstrating a low computational complexity owing to the embedding performed in the spatial domain. We believe that our scheme may be appropriate for authentication and copyright protection of CH images in real-time applications.
KAISER J. GIRI is currently an Associate Professor in computer science at the Islamic University of Science and Technology, Awantipora, Kashmir, India. His research interests include digital signal processing and natural language processing. JAVAID A. SHEIKH (Member, IEEE) is currently an Assistant Professor at the Department of Electronics, University of Kashmir, Srinagar, India. His research interests include image processing, wireless communications, design, development of efficient multiple-input, multiple-output orthogonal frequency-division multiplexing-based wireless communication techniques, spread spectrum modulation, digital signal processing, and electromagnetics.
AMIR H. GANDOMI (Senior Member, IEEE) was an Assistant Professor with the Stevens Institute of Technology, USA, and a Distinguished Research Fellow with the BEACON Center, Michigan State University, USA. He is currently a Professor in data science and an ARC DECRA Fellow with the Faculty of Engineering and Information Technology, University of Technology Sydney. He is also affiliated with Obuda University, Budapest, as a Distinguished Professor. He has published over 330 journal articles and seven books which collectively have been cited over 34000 times (H-index = 85). He has been named as one of the most influential scientific minds and Highly Cited Researcher (top 1 publications and 0.1% researchers) for six consecutive years, from 2017 to 2022. He also ranked 18th in GP bibliography among more than 12000 researchers. His research interests include global optimization and (big) data analytics using machine learning and evolutionary computations in particular. He has served as an associate editor, an editor, and a guest editor in several prestigious journals, such as an Associate Editor of IEEE Network Magazine and IEEE INTERNET OF THINGS JOURNAL. He is active in delivering keynotes and invited talks.
MOHAMMAD HIJJI (Member, IEEE) received the Ph.D. degree in computing from Coventry University, U.K., in July 2017. He was the Chairperson of the Computer Science Department, Faculty of Computers, and Information Technology (FCIT), University of Tabuk, Saudi Arabia, from 2020 to 2022, where he is currently the Vice Dean of Development and Quality at FCIT. His research interests include artificial intelligence, cyber security, the Internet of Things (IoT), smart city, energy optimization, disaster, and emergency management. and an Assistant Professor (Tenure-Track) at the Department of Applied AI, School of Convergence, College of Computing and Informatics, Sungkyunkwan University, Seoul, Republic of Korea. His research interests include intelligent video surveillance, medical image analysis, information security, video summarization, multimedia data analysis, computer vision, the IoT/IoMT, and smart cities. He has registered ten patents and published over 220 papers in peer-reviewed journals and conference proceedings in his research areas. He is an associate editor or an editorial board member of more than 14 journals. According to the Web of Science, he is among the most highly cited researchers, in 2021.