User Acceptance Factors Related to Biometric Recognition Technologies of Examination Attendance in Higher Education: TAM Model

: Identity recognition is inﬂuenced at all educational levels by biometric technology. The invention of facial recognition technology has added new efﬁciencies to the traditional method of tracking student examination attendance. This study aims to determine whether biometric recognition technologies could be utilized to enhance undergraduate examination attendance systems. The study examined the perceptions of ﬁrst-year college students regarding the system’s use of face recognition technologies. Based on the proposed framework, experimental results were obtained by developing and deploying unimodal and multimodal face recognition methods. Using a quasi-practical design with sample groups, undergraduate students’ perceptions of traditional and biometric examination attendance were compared. Adopting the Theory for Reasoned Action and the Technology Acceptance Model, a questionnaire was distributed and analyzed to determine perception factors. The ﬁndings reveal that perceived ease of use, and trust and security signiﬁcantly impact perceived usefulness. It was discovered that perceived usefulness signiﬁcantly affects behavioral intention to use a system. According to the research ﬁndings, multimodal biometric recognition receives signiﬁcantly more positive ratings than unimodal biometric recognition. This study proposes that universities utilize biometric technology, particularly facial recognition, to assess users’ acceptance of the system.


Introduction
Universities have a significant requirement that students participate in class examinations. Student examinations are essential to students' academic success at all educational levels [1,2]. Traditionally, students are required to present identification to proctors before entering the examination room to verify their identities. The signature of each student is required on an examination attendance list. The traditional method of identity verification examines identification cards, driver's licenses, and passports. This method is insufficient in terms of accuracy and efficiency, which can lead to class discipline, monitoring, and fraud [3], especially in a large examination classroom [4]. Student identification is susceptible to forgery, loss, obsolescence, and damage. Students could sign the examination attendance list on behalf of classmates who could not attend. In large examination rooms, proctors may take longer to verify students' identities, which may negatively affect students' cognitive psychology [5]. Due to human error, proctors need technological tools to ensure the accuracy of student verification [6]. A modern university uses information technology extensively in a variety of areas, including admissions, registration, surrounding information, schedule management, learning, exams, and web-based exams [7][8][9][10]. The well-known online systems are made to handle important information and general tasks [11], and students use a username and password to log in and verify their identity [12].
Utilizing biometric technology to verify students is a reasonable solution. A student is not required to carry any identification with them. Numerous biometric techniques are being used at present in various applications, involving fingerprint authentication for check-in and check-out [13] and face recognition for granting access to cloud services [14].
To ensure system accuracy and efficiency, some propose using unimodal and multimodal biometric techniques [15][16][17][18]. At present, no proposed verification system utilizes multimodal biometrics to verify student attendance on examinations. The hard work of figuring out how to use information technology to verify a student's identity is crucial to helping students with exam attendance verification.
Several identification and verification systems based on biometric recognition have been proposed, such as fingerprint [19,20] and facial recognition [21][22][23][24][25][26]. The other proposed verification system made use of barcodes [27], QR codes [28], Near-Field Communication (NFC) [29], and Radio Frequency Identification (RFID) [30][31][32]. Ahmed et al. [33] used fingerprint recognition to establish a security framework for online examination. Emmanuel and Okonkwo [34] also developed a biometric authentication system for Nigerian students. The other reason an educational institution might use a biometric system [35] is to create a biometric system that can use the iris to check if a student is in class.
Biometric technology's advantages, made possible by the Internet of Things (IoT), have resulted in its widespread use as a verification device in various applications. While Awojide et al. [36], Jain et al. [37], and Zainal et al. [20] have proposed an IoT user attendance system based on fingerprint recognition, the issue to date of a high number of users may result in a long queue and require a significant amount of time to verify each student. Facial recognition is an effective technique that requires a large amount of personal data. Numerous studies on unimodal face recognition have been investigated to allow students to verify a student's classroom attendance automatically [26,38].
Our previous work designed and explained a framework and architecture for student biometric recognition [39]. The authors developed an IoT-based student biometric recognition system for examination attendance in this work. As such, this system examines students' perceptions of the examination attendance system using facial biometric technologies compared to traditional and biometric approaches, particularly unimodal and multimodal biometric recognition. There are several critical concerns to be addressed in this study to improve student recognition and simplify the operational practice of facial recognition. The research questions were developed to support the study's objectives as follows: Q1: What affects students' perceptions of using biometric technology for identity verification of examination attendance? Q2: Which biometric technology, based on unimodal or multimodal face recognition, do students desire to use for the purpose of attendance verification during examinations?

Biometric Technology
Biometric technology is widely used for two primary purposes: identification and access control [40]. Numerous biometric technology studies are underway, at present, in various recent areas, including algorithms, architectures, modalities, and empirical studies. Three broad categories of biometric studies exist. Physical biometric verification entails the examination of physical characteristics of the human body, including the face, fingerprints, palm, hand geometry, retina, and iris. Behavioral biometrics studies human operations, such as speech, signature, posture, keystroke dynamics, and other behavioral biometric authentication activities on smartphones [41]. Chemical biometrics is the study of chemical cues associated with humans, such as personal odor. The human face is the most important biometric feature, as it is used in various applications, including felon recognition, security systems, forensics, observation systems, and credit card verification.
Generally, biometric systems are classified into two types [42]: unimodal and multimodal [43,44]. A unimodal system uses a single biometric source, such as the face, iris, fingerprint, palm, or other human body parts. By enhancing the performance of a biometric recognition system known as multimodal biometric recognition, multiple physical characteristics of humans can be identified. Four subcategories of multimodal biometric Sustainability 2023, 15, 3092 3 of 18 recognition systems exist, including (1) multiple modalities: the recognition of a subject using more than two types of biometric technologies, such as a face and fingerprint. Ammour et al. [45] used facial and iris recognition to identify individuals in the system. Gunasekaran et al. [46] identified individuals through face, fingerprint, and iris blending.
(2) Multiple sensors employ the same inspection pattern, for example, by combining images from two cameras. Zhao et al. [47] detected objects using two cameras on a mobile phone.
(3) Multiple features: employ multiple algorithms to extract features from images or data, for instance, fingerprint extraction is to be used in conjunction with the first and second algorithms. (4) Multiple and repeated instances: using one or more biometric forms, such as left-and right-iris images for iris recognition, or using the same biometrics in recognition processes. Ye et al. [25] presented a concise overview of deep learning techniques for the re-identification of individuals. Additionally, numerous recent research papers discussed the multimodal biometric approach. Fenu et al. [21] described a multi-biometric recognition system for continuous learner validation in e-learning systems. The system authenticates and authorizes students' entries using their faces and fingerprint. Traore et al. [24] proposed a framework for an online examination that incorporates a multimodal biometric system that utilizes the continuous capture of images via a web camera.
Numerous published articles on contemporary biometric research discuss multimodal biometric technologies. Chen et al. [48] proposed a multimodal framework in their pioneering Variational Bayesian Extreme Learning Machine (VBELM). They created a block using facial images and fingerprint templates generated from a feature-image matrix. The proposed approach retrieved the core layer semantic aspects of local features, resulting in increased characterization capabilities, element reduction, and improved correctness for multimodal biometrics. In terms of generalization, testing accuracy, efficiency, and stability, the VBELM outperforms traditional methods. Gomez-Barrero et al. [49] described a technique for improving a multimodal biometric approach based on the homomorphic encoding technique, which encrypts all database data. Multimodal biometric fusion is a term that refers to the combination of features, scores, and decision levels. The experiments were performed using an online signature and fingerprint recognition database. The system complies with the ISO/IEC 24745. Biometric technology demonstrates that the biometric approach is applicable in a wide variety of fields. As a result, this study focused on the applications of unimodal and multimodal biometric recognitions that can be used to determine a student's examination attendance in place of the traditional method.

Related Face Recognition
Face recognition is the most widely used method for identifying and verifying individuals in person. A face recognition system's three critical processes are detection, feature extraction, and facial similarity. Face recognition applies to a broad range of research fields. For instance, Dass et al. [50] concentrated on face recognition techniques and real-time face recognition on the Raspberry Pi. Yadav and Vishwakarma [51] proposed a novel, advanced, efficient framework based on interval type-II fuzzy membership and a kernel-based sparse method. They quantified the pixels in a participation image using type-II fuzzy logic and a membership function for type-II extended intervals. When K-nearest neighbor and Euclidean distance metrics were used for sparse representation, the experimental analysis revealed a 10% increase in accuracy. According to Nguyen et al. [52], the human face is a passive biometric. Face images work well when combined with the individual's interaction, including blinking, nodding, and turning the face. The facial recognition system is an artificial intelligence system that identifies individuals through image patterns derived from the textures and shapes of their faces. Temperatures and wrinkles do not affect the visible imagery of facial biometrics. Yaddaden et al. [53] demonstrated the effectiveness of a facial recognition system built on a convolutional neural network architecture and equipped with an error detection module. The experiment used five benchmark facial expression datasets with promising results. For example, the accuracy was higher than 95%, and the number of false positives was cut by 20%.

User Perception Model
The Technology Acceptance Model (TAM) was adopted and developed to examine the use of system acceptance [54]. The principal keys of TAM influenced users' intention to use technology by assembling perceived usefulness and perceived ease of use. BioTAM was a new resolution for a biometric authentication system's technology acceptance model [55]. BioTAM proposed that factors, such as user interfaces, biometric enrollment or verification procedures, devices, and other auxiliary tools, were used to study the end-behavioral user's intention toward using a biometric system. The questionnaire was statistically evaluated based on confounding variables. Wang [56] extended a modified TAM to investigate financial biometric identification applications. The study suggested that perceived privacy and perceived trust were added as modified TAM variables for biometric identification, which includes the face, fingerprint, iris, and voice.
Additionally, TAM is not limited to the study of user perception. Ajzen and Fishbein [57] proposed the Theory for Reasoned Action (TRA) as a framework for forecasting and describing actions across a range of domains, including attitude, intention [58], and motivation. TRA is a very broad concept intended to describe essentially any aspect of human actions. To summarize, the proper determination of user perception is contingent upon the users' intention to use technology. Buabeng-Andoh [59] proposed a model for predicting students' intentions regarding mobile learning adaptation. A combination of TAM and TRA was used to ascertain students' intentions regarding m-learning usage in the classroom. Banga and Pillai [60] discussed the factors that should be considered. Biometric systems for mobile payments should include accuracy, capability, acceptability, cost-effectiveness, and hygiene factors. Behavioral biometrics measure how an activity was carried out rather than the outcome of the activity. For example, the login system is responsive to the speed with which a user enters a username and password. TAM and TRA were used in this study to establish a practical biometric relationship between six significant factors: perceived usefulness and ease of use, as well as users' trust and security, attitudes, intention to use, and actual system use.

Materials and Methods
This research employed the Internet of Things (IoT) and facial recognition to propose a student biometric recognition system for exam attendance. Comparing biometric technology to the conventional method enables the replacement of student identification cards with face recognition technology in examination rooms. Creating a fake student identification card is simple. Each semester, students are subject to traditional verification using student identification cards. During the midterm exam for the spring semester of 2020, a sampling group of 161 first-year students was subjected to the biometric examination attendance system. Students were separated into four classes for digital intelligence and software science. The unimodal biometric recognition system was used in a single examination session to examine student perception. The multimodal biometric recognition system was also tested on the same test subjects during a later examination session.

Proposed Face Recognition Framework
In order to develop a prototype software system for exam attendance, our framework outlined the specifics of biometric student recognition. The extension of the previous framework in this study made possible the deployment and investigation of a student examination attendance system based on the Internet of Things (IoT) in its present state for unimodal biometric recognition with a single camera and multimodal biometric recognition with two cameras, as illustrated in Figure 1. The student examination attendance system ran on a Raspberry Pi 4 with 8GB of RAM and was linked to cameras and a monitor displayed in front of the exam room. Our experiment utilized OpenCV and Python 3.9 to train and test data. There were two primary system users (proctors and students). Teachers and other staff members were assigned as proctors to monitor student examination attendance. A proctor must authenticate with the system and enter the examination session to gain access.  Figure 2a shows the home page of our examination attendance system. As depicted in Figure 2b, each examination session includes information about the examination, including the date, subject, time, student list, and proctors. Before entering the examination session, students must register their faces and personal information in the system. The real-time database system contained the facial templates of students from which the OpenCV face recognition technology-trained system. Prior to entering the examination room, when a student approaches, the proctor uses the verification mode to identify the student's face. The student examination attendance system can detect students' faces, validate their identity accuracy, and report on a student's rate of accurate face recognition.
nition with two cameras, as illustrated in Figure 1. The student examination attendance system ran on a Raspberry Pi 4 with 8GB of RAM and was linked to cameras and a monitor displayed in front of the exam room. Our experiment utilized OpenCV and Python 3.9 to train and test data. There were two primary system users (proctors and students). Teachers and other staff members were assigned as proctors to monitor student examination attendance. A proctor must authenticate with the system and enter the examination session to gain access. Figure 2a shows the home page of our examination attendance system. As depicted in Figure 2b, each examination session includes information about the examination, including the date, subject, time, student list, and proctors. Before entering the examination session, students must register their faces and personal information in the system. The real-time database system contained the facial templates of students from which the OpenCV face recognition technology-trained system. Prior to entering the examination room, when a student approaches, the proctor uses the verification mode to identify the student's face. The student examination attendance system can detect students' faces, validate their identity accuracy, and report on a student's rate of accurate face recognition. In addition, as shown in Figure 2c, a single camera was used to identify and verify the student's face, a technique known as unimodal biometric recognition. Multimodal biometric recognition requires two cameras to identify and verify the student's face; the system then calculates the average matching score, as shown in Figure 2d. Students gain practical experience with a cutting-edge face recognition system. The educational repercussion of biometric technology is the most challenging to implement. Therefore, in this In addition, as shown in Figure 2c, a single camera was used to identify and verify the student's face, a technique known as unimodal biometric recognition. Multimodal biometric recognition requires two cameras to identify and verify the student's face; the system then calculates the average matching score, as shown in Figure 2d. Students gain practical experience with a cutting-edge face recognition system. The educational repercussion of biometric technology is the most challenging to implement. Therefore, in this study, the authors compared how people planned to make decisions based on how they used and satisfied biometric technologies in education.

Instrument
The questionnaire was separated into two major sections. The primary section examined the respondents' demographic characteristics (age and gender) in our experimental study. The secondary section of the study examined student perceptions using the Technology Acceptance Modal (TAM) and Theory for Reasoned Action (TRA). The following five components were assigned to students: perceived usefulness (4 items), perceived ease of use (6 items), trust and security (5 items), attitude (4 items), behavioral intention to use (3 items), and actual system use (2 items). The survey respondents evaluated every item on a 5-point Likert scale, varying from 1 (strongly disagree) to 5 (strongly agree). This study included a total of twenty-three items. After the multimodal biometric verification was completed and the students completed their examinations, the participants were surveyed.
x FOR PEER REVIEW 6 of 18 study, the authors compared how people planned to make decisions based on how they used and satisfied biometric technologies in education.

Instrument
The questionnaire was separated into two major sections. The primary section examined the respondents' demographic characteristics (age and gender) in our experimental study. The secondary section of the study examined student perceptions using the Technology Acceptance Modal (TAM) and Theory for Reasoned Action (TRA). The following five components were assigned to students: perceived usefulness (4 items), perceived ease of use (6 items), trust and security (5 items), attitude (4 items), behavioral intention to use (3 items), and actual system use (2 items). The survey respondents evaluated every item on a 5-point Likert scale, varying from 1 (strongly disagree) to 5 (strongly agree). This study included a total of twenty-three items. After the multimodal biometric verification was completed and the students completed their examinations, the participants were surveyed.

Instrument Validation and Analysis
This study's methodology was quantitative. A questionnaire was initially developed to elicit student perceptions and test hypotheses. Perceptions and hypotheses were developed in accordance with DeVellis and Thorpe's guidelines [61]. The proposed method generated a list of components to evaluate, defined the items within each element, specified the measurement scale, and validated the model's reliability and validity. Cronbach's alpha [62], compound reliability (CR), convergent validity (AVE), and item analysis were used as indicators in the final step. SPSS was used to perform Exploratory Factor Analysis (EFA) and construct validity checks on the biometric recognition processes. Additionally, using the constructs' covariance matrix, the Confirmatory Factor Analysis (CFA) designed on Structural Equation Modeling (SEM) was assessed. Following that, one hundred and sixty-one students were asked to assess the end-user perception of facial biometric recognition with a 100% of response rate.
In addition, all study participants were given informed consent to ensure their con-

Instrument Validation and Analysis
This study's methodology was quantitative. A questionnaire was initially developed to elicit student perceptions and test hypotheses. Perceptions and hypotheses were developed in accordance with DeVellis and Thorpe's guidelines [61]. The proposed method generated a list of components to evaluate, defined the items within each element, specified the measurement scale, and validated the model's reliability and validity. Cronbach's alpha [62], compound reliability (CR), convergent validity (AVE), and item analysis were used as indicators in the final step. SPSS was used to perform Exploratory Factor Analysis (EFA) and construct validity checks on the biometric recognition processes. Additionally, using the constructs' covariance matrix, the Confirmatory Factor Analysis (CFA) designed on Structural Equation Modeling (SEM) was assessed. Following that, one hundred and sixtyone students were asked to assess the end-user perception of facial biometric recognition with a 100% of response rate.
In addition, all study participants were given informed consent to ensure their contribution was anonymized and voluntary. The information presented was especially for the purpose of research. Justice, respect, autonomy, compassion, and confidentiality were all guaranteed as ethical values. thors labeled the following hypothesis about student perceptions of biometric recognition for examination attendance: Perceived ease of use has a positive effect on perceived usefulness (H1); trust and security have a positive effect on attitude (H2); perceived usefulness has a positive effect on behavioral intention to use (H3); perceived ease of use has a positive effect on behavioral intention to use (H4); attitude has a positive effect on behavioral intention to use (H5); trust and security have a positive effect on behavioral intention to use (H6); and behavioral intention to use has a positive effect on actual system use (H8).

Evaluation Design and Data Collection
Using the extended TAM model, this study investigated the factors influencing the acceptance of face recognition technologies. The SEM model validated and verified the proposed hypothesis [63]. According to [64][65][66], at least 200 participants were required for an acceptable SEM analysis or at least 5 cases per parameter for a simple SEM model. Since this study contained 24 observable variables, the minimum sample size was 24 × 5 = 120. Students were used to selecting 161 user samples voluntarily. This research did not involve hazardous chemicals, equipment, procedures, animal or human testing, or the use of animals. However, this study obtained informed consent from individuals before collecting biometric data using and handling their facial biometric data. A consent form for facial biometric collection outlined the introduction, purpose, use, handling, rights, and contact information. As a part of the ethical research, the authors respected the participants' voluntariness, anonymity, freedom, and confidentiality. Students could make an informed decision about providing consent and then accept an individual's consent signature for collecting and using their facial biometric data.

Biometric Data Management
The management of collected biometric data is a crucial aspect of protecting the privacy and safety of individuals. Biometric data must be stored securely, using encryption and other security measures to prevent unauthorized access or data breaches. Administrators and authorized personnel must be the only ones with access to biometric data.

Evaluation Design and Data Collection
Using the extended TAM model, this study investigated the factors influencing the acceptance of face recognition technologies. The SEM model validated and verified the proposed hypothesis [63]. According to [64][65][66], at least 200 participants were required for an acceptable SEM analysis or at least 5 cases per parameter for a simple SEM model. Since this study contained 24 observable variables, the minimum sample size was 24 × 5 = 120. Students were used to selecting 161 user samples voluntarily. This research did not involve hazardous chemicals, equipment, procedures, animal or human testing, or the use of animals. However, this study obtained informed consent from individuals before collecting biometric data using and handling their facial biometric data. A consent form for facial biometric collection outlined the introduction, purpose, use, handling, rights, and contact information. As a part of the ethical research, the authors respected the participants' voluntariness, anonymity, freedom, and confidentiality. Students could make an informed decision about providing consent and then accept an individual's consent signature for collecting and using their facial biometric data.

Biometric Data Management
The management of collected biometric data is a crucial aspect of protecting the privacy and safety of individuals. Biometric data must be stored securely, using encryption and other security measures to prevent unauthorized access or data breaches. Administrators and authorized personnel must be the only ones with access to biometric data. Biometric information should be kept only as long as necessary for the purpose for which it was collected and should not be shared with third parties. When the data are no longer required, they should be deleted or anonymized to protect the privacy of individuals in accordance with all applicable laws and regulations, data protection and privacy laws, and to ensure that individuals have the right to access, correct, or delete their personal information. This study was conducted in accordance with all the applicable data protection laws and regulations, including the General Data Protection Regulation (GDPR) in the European Union, the California Consumer Privacy Act (CCPA) in the United States, and Thailand's Personal Data Protection Act BE 2562. (PDPA).

Result and Discussion
Experiments were used to determine student perceptions of facial recognition for the examination attendance system, followed by a survey of 161 students to collect the data. Students were exposed to the conventional method of identifying students with a student card, unimodal face recognition with a single camera, and multimodal face recognition with two cameras. Each participant completed a survey questionnaire at the conclusion of the three examination sessions. This section provides a summary of the study of the perceptions performed.

Desctiptive Statistic
Males were the majority of participants (57.10%). The ages were 19 years old (33.50%). Table 1 shows the student demographic information in this research. The numerical assessment results for each aspect variable in the survey are shown in Table 2. In all the aspects, the three variables with the highest mean score (in order) were "The student identification using biometric recognition is reliable" (4.04). "The biometric recognition would be physically invasive" and "I would trust the face recognition system" had the same high mean score (4.01). On the other hand, the variables with the lowest mean score (in order) were "The biometric recognition does not require much effort to identify me." (2.74), "The biometric recognition is easy to use" (2.80), and "I think the face recognition for examination attendance is more useful than the traditional method (student card)" (2.84).

Reliability Test
Confirmatory Factor Analysis (CFA) was utilized to validate the reliability test. Hair et al. [73] defined construct validity so that the degree to which a collection of observed variables correctly signified the theoretically quantifiable latent variables. Additionally, they assessed the convergent and discriminant validity of postulated criteria. The findings confirm a total of 23 items. As Bagozzi and Yi [74] suggested, no items were removed because the standardized item loading exceeded 0.5. The factor loadings were more significant than 0.50 and ranged from 0.606 to 0.992, indicating their high reliability. Additionally, the instrument's reliability, consistency, and validity were evaluated. As suggested, the composite reliability (CR) value was greater than 0.70 and fell between 0.773 and 1.000 [75]. The average extracted variance (AVE) was greater than 0.50 and ranged between 0.533 and 0.999, indicating that the data are highly reliable. Cronbach's alpha coefficient values were greater than 0.70 and ranged between 0.774 and 0.932, as Cronbach [62] suggested. As shown in Table 3, all achieved and recommended measures and values demonstrate convergent validity acceptance (CR > 0.70 and AVE > 0.50).  The discriminant's validity was established because the square root of each construct was more significant than their corresponding inter-construct correlation estimates, as illustrated in Table 4. Therefore, all of the constructs were tested for their reliability and validity, which also showed a significant level.

Measurement Model Testing
According to Figure 4, the measurement model confirms that all six constructs (PU, PEU, TS, ATT, BIU, and ASU) are the primary factors in the study that were examined using the EFA model. The authors removed no items from the instrument to achieve the best fit for the measurement model because their regression weights were larger than the bare minimum acceptance measure of 0.60. To validate the measurement model, confirmatory factor analysis was used. The authors tested the model-fit determines against the model's advised fit indicators (x 2 /df, GFI, RMSEA, RMR, NFI, CFI, IFI, and TLI) to validate the overall goodness of the fit index. The results show that all values are significantly greater than their relevant measures' (conventional least acceptance) quantities [73]. The model measurement of suitability indicators is shown in Table 5.

Structural Model Estimation
As shown in Figure 5, our proposed exploration model was evaluated using SEM. Hair et al. [73] explained how exploratory and confirmatory factor analysis techniques could simultaneously examine multiple dependence relations. When the model's constructs have direct and indirect effects on one another, this analysis method is particularly useful. The initial step in interpreting SEM results is to examine the model-fit indicator, which demonstrates that the data perfectly fit the proposed model.
As shown in Table 6, the SEM model demonstrates that all fit indicators meet the bare minimum acceptable standards. Figure 5 depicts the path significances, coefficients, and variances justified for each affected variable. could simultaneously examine multiple dependence relations. When the model's constructs have direct and indirect effects on one another, this analysis method is particularly useful. The initial step in interpreting SEM results is to examine the model-fit indicator, which demonstrates that the data perfectly fit the proposed model. As shown in Table 6, the SEM model demonstrates that all fit indicators meet the bare minimum acceptable standards. Figure 5 depicts the path significances, coefficients, and variances justified for each affected variable.

Discussion
According to our findings from the first research question, perceived usefulness is significantly predicted by trust and security (H8) and perceived ease of use (H1). Our accuracy factor was indicated in our item questions under the trust and security construct. This study is consistent with Sidharta's, Priadana's, and Affandi's [76], and Norfolk's and O'Regan's [72] findings on trust and security based on biometric technology. They asserted that a strong correlation existed between accuracy and perceived usefulness. Kanak and Sogukpinar [55] discovered that perceived ease of use of biometric authentication systems had a significant relationship with perceived usefulness in a situational relationship. Our findings support the notion that students' experiences with biometric technology have a strong correlation with perceived ease of use and usefulness for identifying and verifying examination attendance. Students expressed a concern that using a face recognition system would be more convenient than the traditional approach. In this study, a construct of trust and security (H2) had a positive effect on attitude. However, Norfolk and O'Regan [72] discovered an unexpected relationship between accuracy and attitude toward use. On the other hand, this study discovered that trust and security (H6) had no positive effect on behavioral intention to use. Our findings imply that students involved in biometric technology can be trusted to maintain a high level of security while minimizing identity fraud. For various reasons, students consider the employment of biometric technology as an intelligent tool for examination attendance.
The findings indicate that perceived usefulness (H3) positively affects behavioral intention to use. Additionally, behavioral intention (H7) was associated with a beneficial effect on actual system use, as previously mentioned [77,78]. This evidence demonstrates the numerous benefits of biometric technology for actual biometric recognition systems in universities. Surprisingly, the results show that students' perceptions of perceived ease of use (H4) and attitude (H5) have no positive effect on their behavioral intention to use, which is supported by Norfolk and O'Regan [72].
The results of the second research question indicate that multimodal biometric recognition receives significantly more favorable ratings than unimodal biometric recognition. Student perceptions of perceived usefulness and perceived ease of use received the highest scores compared to other factors in the same constructs. According to Labayen et al. [79], students are positively affected by multimodal biometric recognition.

Theoretical Contributions
Biometric technology has several potential applications, including K-12 education, higher education, teacher education, and training. The empirical findings of this study demonstrate that the application of biometric technology can transform traditional examination approaches, challenge established approaches to biometric examination attendance, and enable exam aspects to improve security, reliability, and efficiency. Additionally, this study discovered that when conservative and biometric identifications were compared, they resulted in significantly different experience outcomes. Students preferred the face recognition system over the conservative approach for examination attendance, which is attributed to concerns about trust and security. Biometric recognition enables students to easily and accurately identify and verify themselves. The results produced the following conclusions as the study's primary contribution:

•
The framework for biometric examination attendance recognition was proposed and a prototype application was developed. The study demonstrates the educational biometric recognition framework's practical outcomes, emphasizing unimodal and multimodal face recognition for first-year undergraduate students. Additionally, the proposed architecture and system provide real-time face recognition of students for examination attendance, information, and accuracy rates of face recognition.

•
To investigate students' actual system use, this research adopted a Model of Technology Acceptance (TAM) and a Theory of Reasoned Action (TRA). The educational biometric recognition factors considered student perceptions of ease of use, usefulness, attitude, trust, and security. • Trust and security are significantly related to IoT-based face recognition for class attendance because they are essential to protect individuals' rights and privacy, ensure compliance with laws and regulations, maintain the integrity and security of the system, and build trust with individuals.

•
In comparison to traditional and biometric recognition for examination attendance, multimodal face recognition is significantly more useful than unimodal face recognition.

Practical Implications
The biometric recognition framework of the examination attendance system can be used to enhance the proctoring approach for student identification and verification in elementary, secondary, and higher education. This study can provide new perspectives and techniques for administrators, educators, and instructors with perception into an innovative biometric recognition system for educational examinations and enhance security, reliability, and efficiency.
At present, users increasingly accept biometric technologies in their daily lives, owing to the convenience and speed with which they can be identified and verified. The authors exclusively focused on face recognition for examination attendance in this study. Biometric recognition can track students' movement between areas and eliminate the need for identification in various other ways, such as no student cards. Since 2019, due to the spread of coronavirus disease, universities worldwide have adopted distance education, considered the best solution for social isolation, less contact, and disease prevention.
Several steps can be taken to prevent errors in recognizing a specific image of a student and any significant changes to their appearance in facial recognition systems used for class attendance. One potential solution is to periodically update the stored images of students in the system to reflect any changes in their appearance. This approach could involve regularly scheduled re-enrollments or updates to the system, where students would be required to provide updated images of their current appearance. Another solution is to use more robust multimodal facial recognition systems that use multiple modalities, such as thermal, depth, and 3D images, to improve the system's accuracy and make it more resistant to changes in appearance. Another approach could be to use a combination of facial recognition technology with other forms of identification, such as a student ID card, a password, or a PIN. To ensure that even if the facial recognition system cannot recognize a student due to changes in their appearance, they will still be able to prove their identity and access the class. It is also important to communicate with students about the potential impact of changes in their appearance on the system's accuracy and to provide guidance on how they can update their data in the system in case of significant changes.

Limitations and Future Work
The limited use of facial recognition technology raises ethical concerns from different perspectives, such as privacy, bias, transparency, surveillance, discrimination, misuse of data, and lack of proper regulations. These ethical concerns must be carefully considered and addressed to ensure that the technology is used responsibly and ethically that respects individuals' rights and dignity. It is important to have clear guidelines, regulations, and oversight mechanisms in place to govern the use of facial recognition technology to minimize any negative impacts.
Future research on IoT-based face recognition for class attendance systems should detail potential areas that can be undertaken to improve accuracy and performance. These include using multimodal systems, deep learning algorithms, data privacy and security, addressing biases in the system, integration with other technologies, and real-time monitoring. Researchers and educators could elucidate additional factors influencing individuals' preferences for biometric recognition in their examination approaches. The university may announce biometric recognition for student examination attendance without obtaining permission. However, university administrators would be sensible to obtain consent from students and research their perceptions before biometric recognition is used to replace traditional processes. Thus, researchers and educators could further elucidate the factors influencing individuals' preference for biometric recognition in their examination approaches. Moreover, student biometric recognition can integrate online learning and assessment systems, ensuring that students remain focused during their study sessions. These trends are likely to persist in subsequent research. Institutional Review Board Statement: Ethical review and approval were waived for this study because this research does not involve hazardous chemicals, equipment, procedures, animal or human testing, or the use of animals and human as a subject experiment. Rangsit University granted ethical approval, and the actual protocol number is RSU-GRAD 537/2561.

Informed Consent Statement:
Informed consent was obtained from all study participants. Participants were at least 18 years old. As part of the ethical research, the authors respect the voluntariness, anonymity, freedom, and confidentiality of the participants. The provided data contained no information that could be used to determine the participants' identities.