3.1 Evolution and productivity

The development and adoption of automated proctoring systems have expanded significantly over time, as shown in the timeline in Fig. 2. Building upon the generalized automated proctoring system shown in Fig. 3, it undergoes continuous enhancement.

Fig. 2

Timeline chart illustrating the progression of online education and the adoption of automated proctoring systems over time

Fig. 3

Generalized automated proctoring system process

The publication trends in a particular subject or field can provide a substantial indication of the research level and enthusiasm within that field. These trends play a crucial role in evaluating the level of maturity within the field. Figure 4 illustrates the Diachronic productivity in the yearly production of articles on automated proctoring systems in the WoS database. The initial works identified date back to 1996. The longitudinal analysis spanning nearly 27 years reveals a progressive increase in article production on the subject, albeit at varying rates.

Our research identified three clearly defined phases: Between 1996 and 2007, researchers published only seven documents, accounting for a mere 2.66% of the total. From 2008 to 2019, the publication rate showed a moderate increase, with the publication of 42 papers, which accounted for 15.96% of the total. The results represented a substantial surge in comparison to the preceding period. During the third phase, which lasted from 2019 to 2023, there was a significant increase in growth, evidenced by the publication of 220 scientific papers. These papers constitute 83.65% of the total, demonstrating a remarkable rise compared to the previous two phases. Researchers published 50 documents in both 2020 and 2021. Price’s law posits that the growth of scientific information is exponential in nature, indicating rapid and significant expansion. Consequently, over a span of 10–15 years, this leads to a twofold increase in scientific literature.

Fig. 4

Figure 4 provides evidence for Price’s law, showing a significant increase in publications over a 10-year period (2013–2023), from 25 in 2013 to 237 in 2023. There has been a recent surge in document volume, potentially attributed to ongoing factors such as the COVID-19 pandemic and the transition to online education. As a result, technological advancements in educational systems have gained greater importance than before.

3.2 General characteristics of publication trends

We obtained 263 documents from the WOS Core Collection Database comprising 217 sources. Figure 5 presents the summary results of these documents as an informative dashboard. The research on online proctoring involved 923 authors, but only 27 authors independently authored the articles. On average, each article had 3.68 authors, and 14.07% of the articles involved international collaborations. The average age of the documents was 5.08. This information yields valuable insights into the trajectory of research within the field, facilitating the development of future studies and guiding strategic research directions.

Fig. 53.3 Characteristics of publications

The research documents about online proctoring were categorized into six distinct groups: article, review, proceeding paper, book chapters, early access, and editorial material, as depicted in Fig. 6, with each category contributing a specific percentage of the total documents. In particular, the” Article” category held the majority with 55.1%, followed by the “Proceedings Paper” category with 35.4%. Book chapters made up the least amount of the remaining percentages (0.4%), followed by editorial pieces (2.7%), reviews (4.0%), and early access materials (3.4%). Compared to other fields, research in the online proctoring domain had a higher proportion of “articles” and “proceedings papers” indicating a greater emphasis on these categories.

Fig. 6

Characteristic description of documents

3.4 Distribution of research directions

The research trajectory diagram shown in Fig. 7 displays 20 current analyses of research areas and focuses on proctoring systems depicting potential pathways and areas of focus for future research [5]. Education and educational research, computer science and information systems, and engineering are the top three fields in proctoring systems. This is because technology has captured much attention in the education sector. Moreover, education research has shifted towards online education and maintaining the authenticity of online exams. Education and educational research provide a humancentric understanding of proctoring’s impact and ethical concerns [19]. On the other hand, computer science and information systems bring technological expertise to build secure and efficient proctoring systems. The collaboration of education and computer science domains indicates that both fields work together to ensure a reliable and trustworthy online exam environment.

Fig. 7

Analysis of research areas and focus

3.5 Dispersion of scientific literature

A law can facilitate the formulation of a theory to elucidate the underlying reasons for the existence of a particular pattern. One such established principle is Bradford’s Law. Bradford’s Law of Scattering Principle is useful for assessing and ranking scientific journals according to the distribution of information resources in scientific literature. This principle organizes journals in a specific field based on the number of articles they publish on a particular topic, with the journals being arranged in decreasing order, as depicted in Fig. 8. The analysis of the scatter of articles in Fig. 6 reveals that online proctoring has garnered significant scholarly attention, with 217 source titles in the form of journals, books, or conference volumes publishing research papers on the subject. The Bradford’s Law graph shows a logarithmic scale on the x-axis (source log-rank) against the number of articles on the y-axis. The core sources publish most articles on a topic, while numerous other sources distribute the rest. The top third of journals, referred to as the core zone, comprises a few highly productive journals that cover a substantial portion of the articles on automated proctoring [20].

Fig. 83.6 Scientific production in top journals

Journals serve as the primary platform for disseminating research papers, with authoritative journals often attracting more rigorous and innovative submissions. Table 4 shows the top ten journals in the field of online proctoring based on the number of published articles. Among those journals, the Journal of Chemical Education published five articles, significantly contributing to the expansion and development of online proctoring studies. This journal has a 2-year impact factor for 2022 of 3.0, 19,403 citations, and a CiteScore of 5.2 for 2022. Six other journals published four articles each. The data distribution is right-skewed, indicating that most of the publications have a low number of articles.

Table 4 The top 10 journals that include proctoring documents indexed in WoS

In contrast, a few publications have a significantly higher number of articles. The Emerging Sources Citation Index (ESCI), part of the Clarivate Web of Science Core Collection, incorporates two top journals: The International Journal of Advanced Computer Science and Applications and the International Review of Research in Open and Distributed Learning [31]. These journals include high-quality, peer-reviewed publications, and though they do not receive impact factors, citations from the ESCI contribute to the impact factors of other journals. Indexing provides a mark of quality and increases the journals’ visibility [32]. Selection criteria for ESCI may be less stringent than for established indexes like SCIE. The top 10 index journals include two SCIE, seven SSCI, two ESCI, and one AHCI, respectively. Most of the leading journals publishing articles related to online proctoring were in the Education & Educational Research category, with other subject categories being Special Education and Social Sciences.

3.7 Cluster analysis

In order to examine extensive collections of bibliographic data, clusters of keywords were created to detect patterns, trends, and connections within the specific area of research. Based on the themes present in the clusters, the clusters are labeled into five potential names, as depicted in Table 5. The data presented in Table 6 keywords found in Cluster 1, including face detection and recognition, machine learning, and security. Cluster 2 highlights the difficulties associated with academic integrity, including stress and its impact on academic achievement. The designated term “COVID-19” is Cluster 3, which illustrates the transformations in the field of education as a result of external influences, such as pandemics. Cluster 4 emphasizes the utilization of technology-driven approaches to tackle the issues of cheating and plagiarism. Cohort 5 underscores the impact of psychosocial variables on both student experience and academic achievement.

Table 5 Potential cluster names

To further strengthen the investigation, we employ a network diagram as shown in Fig. 9 to depict the associations between clusters, both within and between clusters. The network diagram suggests that test anxiety is a challenge in online education. It also indicates that there are security concerns such as cheating and plagiarism.

Fig. 93.8 Keyword analysis

Over time, the keyword co-occurrence knowledge graph demonstrates the relevance of particular topics, and keywords frequently appearing within a specific period and experiencing a significant rise in citations are regarded as indicators of cutting-edge subjects or emerging trends. Keywords serve as essential elements that convey crucial information about a publication and elucidate the extent of regional research within any domain of knowledge [17]. Therefore, it offers a quantitative overview of the research landscape, facilitating the identification of influential topics within the domain. Analyzed using the VOSviewer software, a keyword analysis was conducted on the abstracts of the primary literature regarding automated proctoring in the “WoS” database as shown in Fig. 10.

Among all included keywords, the most frequently mentioned keywords were “performance” (n = 14), followed by “impact” (n = 12), “education” (n = 10), and “student” (n = 10) and “perception” (n = 8). The central position also suggests that these key topics are highly interconnected with other terms in the network. The colors represent distinct clusters or thematic categories, with terms in blue the earliest and in yellow representing the most recent and trending terms. Terms associated with “online learning” are categorized, encompassing concepts such as “e-learning”, “online courses”, “distance learning” and so on. The magnitude of the line’s thickness typically reflects the intensity of the correlation between the terms. A greater line thickness indicates a greater frequency of co-occurrence, suggesting that these topics are frequently discussed in the literature. Distinct hues correspond to distinct knowledge domains grouped by the software’s clustering algorithm. The greater the distance, the weaker the connection between them. The timeline depicted at the bottom of the visualization illustrates the alterations that have occurred between the years 2018 and 2022. The prominence of certain terms has changed over time due to the influence of COVID-19 on education [33].

Fig. 10

Density co-occurrence visualization of keyword terms

3.9 Factorial analysis

In this research, we utilized a dimensionality reduction method known as Multiple Correspondence Analysis (MCA) to examine and enhance prevalent terms in the set of articles [34]. MCA is a survey analysis tool that visually depicts data similarities and relationships between categories, focusing on associations between these categories. It makes a two-dimensional plot with a point cloud of indicator values that are close to each other and have similar properties. The closer the article’s keywords are to each other, the more related they are [35]. Moreover, the closer the indicator’s position is to the map’s center, the greater the number of articles researching the topic within the field. Clustered keywords with high similarity are presented in Fig. 11. Higher positive values in Dimension 1 and Dimension 2 are indicators of a positive association with the underlying structure captured by the dimensionality reduction technique. Cluster 1 variables are related to online learning and assessment, while Cluster 2 contains variables related to traditional assessment and security.

From the analysis, keywords such as “COVID-19”, “remote proctoring”, “assessment”, “online exams” and “academic integrity” have higher positive values in at least one of the dimensions. Education and technology have the highest positive coordinates on the first dimension, indicating their strong association and positive relationship to the first dimension. Stress has the most negative correlation with the first dimension, indicating its negative association with the first dimension and positively associated variables, like education and technology [36]. On the second dimension, variables with positive associations include exams, testing, proctoring, and academic performance [37]. These variables are likely related to the concepts of assessment and testing. Evaluation and security have the highest positive coordinates on the second dimension, suggesting their strong association and a positive relationship to the second dimension. Face detection and biometrics have the most negative coordinates on the second dimension, suggesting their negative association with the second dimension and positively associated variables, like evaluation and security [38]. Other variables with negative associations in the first dimension include cheating, academic dishonesty, and online assessment. These variables are likely related to concerns about academic integrity in online learning environments. On the second dimension, variables with negative associations include online courses, student experience, and teaching, which are related to potential downsides of online learning, such as reduced student engagement or teacher effectiveness [39].

Fig. 11

Factorial analysis: MCA method

3.10 Word cloud

The visual, intuitive representation of text data, referred to as a word cloud, aids in identifying trends and patterns, relationships, comparative analysis, exploratory data analysis, and highlighting focus areas. Figure 12 showcases a word cloud resulting from our study on automated proctoring. The most commonly occurring terms outlay are “performance”, “education”, “students”, “perspectives”, “academic dishonesty”, “formative assessment”, “higher education”, “medical education”, “test anxiety”, “test assessment”, “computer-based assessment”, “experience”, “diversity” and “inclusion”. The word cloud analysis suggests that the study focuses on assessing students’ performance in education through various means, with a particular emphasis on academic dishonesty and using technology in assessment. Additionally, the presence of “diversity” and “inclusion” in the word cloud indicates that many researchers are focusing on making evaluations more fair and equitable for all students. Based on the word cloud study, assessments should accurately measure student learning by focusing on specific findings such as performance [40].

Fig. 123.11 Density visualization using heatmap

We use a density visualization heatmap, a two-dimensional representation of the point concentration of data, to identify regions of high and low activity. The heatmap represents how much focus is given to each term in the research field. The most intense color terms illustrated in Fig. 13 possess the highest density. Assessment, online education, students, online examinations, COVID-19, e-learning, and academic performance. These elements comprise the core subjects of the investigated domain. The term “online education” is associated with concepts such as “students”, “online examinations,” and “assessment” indicating the prevalence of online assessments in online education.

Fig. 13

Density visualization using heatmap

Furthermore, the map illustrates the close relationship between e-learning, students, online examinations, performance, and assessment in the context of online learning [41]. Online examinations, a crucial component of online education, enable the evaluation of student performance in e-learning environments. Examining the efficacy of these assessments and strengthening the proctoring system is a critical area of research [42].

3.12 Hierarchical relationships of topic via dendrogram

The Hierarchical relationship of bibliometric research is given in Fig. 14. Items that are grouped at lower levels of the dendrogram have more in common than items in other clusters. Topics such as “performance”, “academic performance”, “academic dishonesty”, “success”, “satisfaction”, “test anxiety”, “stress”, “exams”, “formative assessment”, “online exams” and “computer-based assessment” are closely related, as they cluster together in the analysis. This suggests they share common variance and measure similar constructs related to academic performance, assessment, and stress factors. “Technology”, “computer”, “technology acceptance”, and “user acceptance” cluster together, suggesting they are related to technology usage and acceptance in educational settings.

Fig. 14

Hierarchical topic tree of bibliometric research

3.13 Leading countries and institutions3.13.1 Country scientific production analysis

The development of onlineproctoring has been enriched by the contributions of scholars hailing from 59 countries. Notably, the United States, China, and India are the most active countries in this space, as evidenced by their scientific publications. As depicted in Figs. 15 and 16, these countries are home to numerous research institutions and universities, according to their prolific production of scientific literature. However, a significant chasm exists between the number of publications produced by these three nations and those from other countries. This disparity has steadily increased in recent years, as the United States, China, and India have invested copiously in research and development, thus indicating a vibrant research community.

Fig. 15

Choropleth map of the top contributing countries to the production of online proctoring scientific literature by the number of work

Fig. 16

Countries’ publications over time

3.14 Country production over time

According to the data presented in Fig. 17 and 59 countries published scholarly-researched articles between 1996 and 2023. The term “country” in this section is used to refer to the country of the institution affiliated with the first author. Notably, the authors of these articles come from various countries. Figure 17 clearly demonstrates that production in the USA remained relatively stable until approximately 2015, when there was a discernible surge. By 2023, production in the USA will have experienced a substantial increase, surpassing that of the other countries mentioned, indicating a long research span for automated proctoring in the USA. According to the top 10 contributing countries, as shown in Table 7, The United States (US) has published the highest number of articles (127), taking 23.43% of the total production of literature, thus securing the top position. Germany follows with 41 articles, the United Kingdom with 48, Turkey with 29, and India with 27.

Fig. 17Table 7 The top 10 contributing countries3.15 Most cited countries

The highest number of citations for an article generally signifies several critical aspects related to its impact, relevance, and recognition; the availability of the journal or platform where the article is published is an essential factor. As shown in Fig. 17, the United States receives the most citations (n = 619), followed by India (n = 124) and Spain (n = 114). Out of the top 10 countries, only India and Pakistan were developing countries [43]. These findings suggest that developed countries have a more prominent role in proctoring systems due to the broader availability of technological advancements. Conversely, research in developing countries lags behind the trend in developed countries.

3.16 Dimensions of the corresponding author’s countries

Single Country Publications (SCPs) reflect research that is primarily concentrated within a single country’s academic community. Multiple Country Publications (MCPs), on the other hand, represent international collaboration and knowledge exchange across borders. The dimensions of the corresponding author countries, as represented in Fig. 18, provide valuable insights into re-search output and collaboration patterns. Notably, Indonesia (93.75%), China (93.33%), and the USA (91.80%) exhibit the highest proportions of SCPs, indicating a strong emphasis on domestic research initiatives. Conversely, Canada boasts the highest proportion of MCPs at 42.86%, followed by Spain (25.00%) and Turkey (18.18%). These patterns provide crucial information about research strategies, funding decisions, and initiatives to promote collaboration.

Fig. 18

SCP and MCP dimensions of the corresponding author’s countries

The R programming language is used to call the bibliometric package to analyze cooperation among countries and get cooperation data between countries [44]. Figure 19 has a choropleth map generated from Biblioshiny, which depicts bibliometric features. The number of publications in a country or region is a delicate indicator. This reflects the degree of concern and research prowess of a country. In this respect, it is worth noting that there is strong collaboration among China, the United States, and the United Kingdom for scholarly literature. The lines across the map illustrate collaborative relationships between countries and regions. The darker the country’s color on the map, the closer the cooperation was between those countries, and vice versa. Figure 20 depict the degree of connections, indicating their influence on collaboration.

Fig. 19

Collaborative relationships in countries and regions

Fig. 203.17 Authors with high impact according to H-Index

The author with the highest productivity is shown in Fig. 21 in order form according to the impact factor concerning the h-index. The ranking shows Atoum Y to have the most contributions.

The main affiliations of institutions can be seen in Fig. 22, which shows the University of Toronto as the most productive, with seven papers published in the analyzed dataset. The University of London ranks second with six papers.

Fig. 21

Most relevant affiliations

Fig. 22

Author co-citation network

3.18 Co-citation analysis

Co-citation analysis often examines journal co-citation, author co-citation, and document co-citation. The size of the nodes in the co-citation network is indicative of the frequency of citations for the journal, author, or publication. On the other hand, the thickness of the edges signifies the frequency of co-citation betwe