Indroduction
Machine learning (ML) is a pioneering area of research in artificial intelligence. ML has a broad application area, from running search engines to protecting passwords. ML is the study of machines learning from human behavior to solve simple to complex problems (Lv & Tang, 2011). Machine learning algorithms revolutionize real-world applications and research directions. Machine learning makes systems more intelligent and automated to handle high-dimensional and complex data (Sarker, 2021). ML is adept at knowledge acquisition, learning, predicting, and solving problems. In real-time, ML saves the time and efforts of every human. It has become inevitable for every human to make their everyday lives smoother from multiple perspectives. This article discusses the role of ML in the academic area. During the COVID pandemic crisis, the agony of students and the hardships faced by education are many. The offline classes were not possible due to widespread virus threats (Tarik, Aissa & Yousef, 2021). The protection of students’ health and safety was a priority. The educational institutions are forced to cancel physical learning sessions in a classroom environment, and conducting examinations is challenging. E-learning, remote learning, and online education became a solution as it’s the only way out for continuous learning.
The examinations or assessments were considered necessary to be conducted online mode. The online education mode came to rescue the aid of students in completing their courses without a break in academia. The exam management system is the most affected in academia during the lockdown scenario. ML played a significant role in paving students toward their academic completion with examinations. The student data are a prerequisite for any analysis related to the academic perspective. The data mining of the student’s data of respective classes and research helps the educational infrastructure to develop a support system for the students and teachers in preparation for final assessments.
The student learning management data contribute to designing a decision-making model for prediction or analysis in the educational domain (Lavoué et al., 2017). Machine learning approaches are the key to the development and process of these learning and assessment models. However, the sample student data should be adequate and informational. The ML models are beneficial in the prediction analysis to study student performance based on the data. This futuristic prediction characteristic of ML facilitates improving student performance and provides early intervention in learning and achieving success in examinations (Sandra, Lumbangaol & Matsuo, 2021). The main advantage of the ML model is its capacity for accurate prediction with limited student data resources. The contribution of ML models proved to be substantial in student academic performance predictions, with learning outcomes aiding in preparation for upcoming examinations (Balaji et al., 2021). The unique ability to explain and interpret enables easy adaptation and operation for students, teachers, and administration. ML requires basic skills and primitive knowledge for its operation that can be implemented in education (Linardatos, Papastefanopoulos & Kotsiantis, 2020).
There is clear evidence that ML is a paradigm shift in the education field to be innovative, interactive, and personalized for collaborative learning. Whether online or offline learning mode, ML has proved to be valuable in outcome-based education for professional or career perspectives of students. The teaching strategy has been shifted to outcome-based education; the academic curriculum has been redefined based on program and course outcomes. For this outcome-based model, the prediction is the basis for building up the curriculum for student assessment preparation and evaluation. There is a demand for constant observation and frequent measurement of student learning ability before appearing examinations. The students’ examinations are centered on these curriculums and Program Educational Outcomes (Yanes et al., 2020). ML serves as an exoskeleton for the instructor to plan the instruction and evaluation models. ML provides teachers a platform for multifaceted presentation, simulation, and automated student feedback. ML bridges the gap between students and teachers for more interactive learning and proximal developments (August & Tsaima, 2021). This article is intended to discuss the impact of machine learning on an event of the learning process, “Examinations”. The key features of this review paper are the role of ML in the examination process during the lockdown and its emergence, multifaceted role in student assessment.
With the current onset of pandemic situation, educational institutions, industries, government organizations, etc. have shifted their inclination towards online mode of teaching/learning, communication and other interactions. In such scenario, there is a high probability of false-doing such as malpractices in examinations and other negative aspects. Hence, automated systems for online interactions must be well-equipped to overcome these drawbacks. The rationale behind this review is mainly focused on this concept by highlighting the merits and demerits of various ML models which can assist in the variety of online applications.
This review is specifically intended for educational institutions who carry out teaching and learning in the online mode. Proper selection of the ML based automated system will enhance the efficiency of online learning, online examinations, etc.. Government organizations which conduct nation-wide/state-wide online examinations also will benefit out of this review.
The motives of this work are:
In artificial intelligence, Machine Learning is the study of how machines can learn human behavior and imitate it for decision making and solving complex problems. The machine learning algorithms started to evolve in 1970 though the term was coined in 1959. The main classification is based on their learning and testing the validity of the proposed model behavior as to whether supervised or unsupervised (Louridas & Ebert, 2016). In supervised learning, classification and regression algorithms utilize training data for prediction. The prediction is validated by the test set data. On the contrary, the machine must find its solutions using the training data in unsupervised learning. The general machine learning approaches and their classification with their techniques are displayed below in Fig. 1.
Enlarge this image.Figure 1: General classification of Machine Learning. DOI: 10.7717/peerjcs.986/fig-1
Search Strategy and Research Questions Identification of research gap
Machine Learning is a vast area with multiple applications in different fields. This review is based on its application in the field of academia. When the generalized survey on ML in education was performed, several papers exhibited the role of ML in learning and the prediction of student performance. However, there was limited focal point review on the extensive role of ML in exam management. Systematic exam management is the need of the hour under lockdown situations. The ML is pivotal in automation and efficient in monitoring and regulation of a sequential examination process. If ML is not implemented the examination process, it may tend to become complicated with huge student data analysis. The manual timetable process can be challenging and may consume more human efforts and time. This inadequacy of work on ML in exam management systems motivated the authors to revisit the papers for the search strategy. This paper tracks the importance of ML from beginning to end of the examinations as an entire cycle. This review paper is structured as a sequence of activities aiming for a complete understanding of the pivotal role of ML in Exam Management Systems, from preparation to assessment and grading.
Search strategy
The motivation of this work is to elucidate empirical evidence supporting the effectiveness of machine learning on online preparation, assessment, and evaluation of student performance. For this review, an extensive search on machine learning was performed. The papers are selected based on recognized and accredited publications such as IEEE, Springer, Elsevier, Wiley, ACM, and International Journals. The articles are indexed in Scopus or web of sciences available in research databases. Machine learning is a popular topic in exhibits and conferences. The recent conference papers with many citations and novel ideas are also included in this study. The specific documents with machine learning keywords are retrieved, ensuring practical relevance in selecting each article. About 140 papers were selected from the last five years (some exceptional previous papers were also included). The search strategy is displayed below in Fig. 2.
Enlarge this image.Figure 2: Search strategy for ML in the exam management system. DOI: 10.7717/peerjcs.986/fig-2
Research question formulation
The following research questions are based on the segmentation of the LEMS:
The above questions are formulated, and this leads into the following divisions: the significance of ML in current lockdown, Pre-Exam phase or preparation phase, Examination phase, Post examination phase (Evaluation and grading), and issues and challenges are discussed, followed by a conclusion. The distribution of retrieved papers based on years and sources are given in Figs. 3 and 4. For LEMS, most of the articles are chosen between 2019–2022, which has a bearing on the lockdown era.
The Emergence of ML in Lockdown Exam Management Systems
There is an unmistakable emergence of the use of Machine Learning tools in the education sector during pandemic times. The contemporary research supports that ML has been a powerful tool implied in learning and assessment during these pandemic times. There is a profound shift in behavior for every student from learning in the classroom to online mode. The pandemic crisis has increased students’ emotional sensitivity, they are in anguish, and there is a big question mark about their future careers (Akour et al., 2021). In several countries, the ministry of education has suspended regular school or college. Face-to-face studies and group education were barred as a precaution. Distance education mode and assessment became obligatory at all levels, from kindergarten school to a doctorate or scholar defense exam (Tarik, Aissa & Yousef, 2021). The personal face-to-face contact to identify the student’s attentiveness and make interventions on the spot is dispensed. The performance prediction can be synchronous; the academic interventions online necessitate monitoring the asynchronous activity of the individual student.
The Machine Learning methods are experts in monitoring. It is also used to develop models to predict students at risk in advance to get the same benefit as in real-time (Karalar, Kapucu & Gürüler, 2021). The ML lends its hand the extra mile to the teachers for smart attendance, teaching, and conducting assessments. The ML made virtual classrooms and assessments possible; the survey was conducted with students for determining student satisfaction during emergency remote learning using Machine Learning. The results indicated that the students favor remote education and assessments (Ho, Cheong & Weldon, 2021). Academic growth is highly impacted due to COVID. The COVID brought a severe setback and sudden change in the modus operandi of academic structure. There are consequences due to COVID, affecting students’ academic standards and behavior to embrace new changes with fear of the future. The ML has been a silent hero that helped the field of academics and assessment to continue with the same integrity (Agarwal et al., 2021). The ML lays building blocks for students to have a safer world and stay ahead amid setbacks.
The examination cycle in the lockdown exam management system (LEMS) has four significant steps
The exam cycle and its subareas focused on this article are illustrated in Fig. 5.
Role of ML in Pre-exams (Preparation Phase) Adaptive learning before examination
Adaptive learning is an approach where the study materials, teaching methods, and inculcation are tuned with the proactive curriculum. The individual learning path is designed for the maximum enrichment of the student’s knowledge and thus enhances the student’s academic achievement. ML must identify the students who are low-level performers to pay personal attention. The ML boosts the student to overcome his vulnerability in the learning process through tailor-made learning modes. However, each student has their behaviors in learning attitude and continual monitoring (Embarak, 2021). ML has a role in designing web-based learning systems by an interface. The learning interface must be adaptive and satisfy individual learner criteria for creating personalized learning. The downside is the non-availability of a collaborative learning approach that has interactions for more efficiency (Oboko et al., 2016). The concepts of brain dominance and students’ psychological behavior serve as input for ML-based models. There is a prerequisite of catering the personalized education to the individual student’s aspiration. This psychological behavior-based ML concept is proposed and tested in practical conditions (Oke et al., 2016).
Modern technological developments demand an intelligent learning environment for students. The teaching strategies should be adaptive to individual student learning characteristics. The ML models enlighten more on personalized adaptive learning for effective pedagogy. The unique student characteristics and personal development are its prime criteria in designing education-based ML models (Peng, Ma & Spector, 2019). Recently, AI-based learning has been prominent in implementing adaptive learning. The systematic mapping method of AI is beneficial for education purposes. The review clearly shows that ML with AI dawns a new era for adaptive learning, helping students in knowledge enhancement (Kabudi, Pappas & Olsen, 2021). The research revisits challenges and issues with the widespread implementation of adaptive learning. The frequent redesigning of curriculum, loss of control in the hands of the faculties, and reluctance to cope with the learning of new tools for updating technology are the issues to be addressed (Mirata et al., 2020).
Predictive analysis of student performance
Assessing a student’s performance at strategic time intervals before exams gives an impetus for improved teaching methods that assist students in achieving success in their exams. The various ML methods and algorithms are extensively used to predict performance in many student communities. The details are inferred below in Table 1.
| Paper ID | Student dataset | ML technique used or compared | Inferences |
|---|---|---|---|
| Kotsiantis, Pierrakeas & Pintelas (2004) | University students for distance learning | (Naïve Bayes) NB, 3-NN, RIPPER | Early predictions are accurate using NB |
| Koutina & Kermanidis (2011) | Postgraduate students | NB, 1-NN, Random Forest (RF), and SMO | The combination of NB and 1-NN is efficient with sampling |
| Sweeney et al. (2016) | Public University | Factorization Machines (FMs), Random Forests (RFs), and the Personalized Linear Multiple Regression (PLMR) | A hybrid and efficient method of RF and FM is proposed. |
| Xu, Moon & Van der Schaar (2017) | Degree students | A novel method is proposed (latent factor model-based course clustering method) | A progressive prediction architecture is proposed for evolving performances. |
| Fagbola et al. (2019) | Programming courses | Linear Regression (LR), M5P decision tree | Designed and evaluated Mobile Interface with ML statistical models |
| Hussain et al. (2018) | Learning session data | LR, ANN, SVM, NBC, and DT | The SVM and ANN models are accurate |
| Sekeroglu, Dimililer & Tuncal (2019) | Secondary school | Backpropagation (BP), Support Vector Regression (SVR), and Long-Short Term Memory (LSTM), | SVR has the highest prediction rate accuracy. |
| Jain & Solanki (2019) | Multiclass students | DT, RF, Gradient Boost | RF outperforms for correlative analysis of student performance |
| Alshabandar et al. (2020) | Massive open online courses | Root Mean Square Error (RMSE) and R-squared | Pass or Fail analysis. But lack of inclusion of temporal features. |
| Shashi, Sunil & Kumkum (2020) | University students | e Logistics Regression, Naïve Bayes, K-Nearest Neighbor, Decision Tree (DT), Support Vector machine | DT outperforms the acquired dataset. |
| Dhilipan et al. (2021) | College students | Binomial logical regression, Decision tree, Entropy, and KNN | Binomial Logical regression is accurate. |
| Pallathadka et al. (2021) | UCI machinery student dataset | Naive Bayes, ID3, C4.5, and SVM | SVM has more accuracy and less error rate |
| Li & Liu (2021) | High school students | CNN, RNN, and DNN | DNN is accurate for a vast dataset |
| Alnassar et al. (2021) | Virtual learning environment | Support Vector Classifier (SVC), k-Nearest Neighbor (k-NN), Artificial Neural Network (ANN) | K-NN algorithm is more suitable for critical analysis. |
| Baashar et al. (2022) | Higher education | Various ANN algorithms | ANN with data mining methods is more prominently used in higher education |
| Albreiki, Zaki & Alashwal (2021) | Multiple datasets | SVM, DT, NB, KNN | There is a need for dynamic |
| Priya, Ankit & Divyansh (2021) | Secondary schools | Logistic regression, ANN, SVM | Logistic regression performs better in the prediction |
| Hussain & Khan (2021) | Secondary and intermediate levels | DT, GA based DT, KNN, GA found KNN | GA based DT outperforms all the other methods |
| Xiao, Ji & Hu (2021) | Survey on different students | DT, NB, MLP, RF, SVM, KNN | Shortcomings of each prediction method were discussed |
| Qiu et al. (2022) | E-learning environment | SVM, NB, DT | A behavior-based prediction model is designed and evaluated. |
DOI: 10.7717/peerjcs.986/table-1
Identification of at-risk students before examinations
The ML has a significant role in predicting forthcoming performance. Besides that, it is also proficient in identifying vulnerable students (low achievers) and a risk category of failure in exams. This prediction and student identification aid the teachers in designing a practical pedagogy for developing personalized monitoring and individual attention. ML facilitates more students to achieve success in coming examinations. The generalized model to predict any at-risk students involves five mandatory steps, as shown in Fig. 6.
The student dataset is preprocessed in an at-risk model, and the required features are extracted. The features such as academic, demographic, social and behavioral factors of the students in prior semester are considered for data set creation as input into ML model for at-risk identification. The ML models are based on algorithms such as Naïve Bayes, Decision Trees, Random Forest, Neural Networks, and even hybrid methods (Marwaha & Singla, 2019). ML with data snapshot analysis helps find underperforming students who need more attention in programming classes. The students can have rehearsal tasks for more understanding of the concepts (Ahadi et al., 2015). The students’ data of periodic assignments submission helps the ML model predict the students who are prone to risk in the examination (Falkner & Falkner, 2012). When applied to the actual log of Learning Management Systems (LMS), the ML methods can determine up to forty percent accuracy in the early detection of at-risk students. The casual students’ behavior, such as their e-book reading habits with an appropriate ML model, can predict students’ at-risk using classifiers (Chen et al., 2021). By closely monitoring ML techniques output, the instructors can intervene with the student and offer additional support and more adaptive teaching (Kondo, Okubo & Hatanaka, 2017).
The ML also helps the university administration identify at-risk students before examinations. ML resolves the issues of rising university attrition. ML models applied with the administrative data give the university administration a predicting insight of students who may discontinue in the middle of the course and take necessary steps to prevent attrition (Berens et al., 2018). In a virtual learning environment of college education, the ML methods identify at-risk students and even marginal students before the examination. More success for these marginal students can be achieved by providing due attention to the examination (Chui et al., 2020).
Enlarge this image.Figure 6: Steps for identification of at-risk students. DOI: 10.7717/peerjcs.986/fig-6
In massive online open courses (MOOC), ML predictions are more efficient in identifying vulnerable students and improving these students’ performance with ML-based learning achievement mode (Al-Shabandar et al., 2019). Sometimes, manual methods fail to identify some demographic or financial issues that may cause students to be at-risk before examinations. ML methods are pragmatic and more potent in determining the student’s at-risk conditions, which manual processes overlook (Soobramoney & Singh, 2019). There are instances where the socio-economic situation leads a student towards academic desertion before completion. There has been a forty to seventy percent increase in at-risk students in countries like South America. The ML was applied to such scenarios for detecting students and focused on these students’ issues on a national scale (Zea, Reina & Molano, 2019).
The ML has proved its extreme impact on the student population, from university students to the K-12 classes, which determines a career path. The prediction promotes the scholastic achievement of such altering their career path (Li et al., 2020). The ML models can be implemented in levels for more accuracy in prediction. A model with three levels of classifiers can achieve eighty percent of accuracy in prediction (Latif, XianWen & Wang, 2021). The ML model’s prediction is based on static data, but they can even stack the students’ weekly performances and their growth in the academic pursuit and bring out a few at risk. It is a repeated measure process that makes it possible to identify the at-risk students throughout the academic year, not a one-time event (Koutcheme et al., 2022). Whether it is its online courses or any virtual learning environment, it is established that there is evidence of the inevitable role of ML, dedicated to the design of prediction models for at-risk students (Adnan et al., 2021). The ML analyzes precisely and competently, helping the students to receive support and proper mentoring from teachers or to avoid dropping out of examinations.
Monitoring student attendance and learning
The examination may be a pivotal event in every student’s life; the preparation process is necessary. The student’s regular attendance in the virtual class and attentiveness is obligatory for preparation and appearance for examination. Most schools and universities regulate a mandated attendance percentage for appearing examinations. Machine learning can be implemented for face detection and face recognition for monitoring intelligent attendance systems. The low-cost ML method of the camera model is all ready to take photos of the students fifteen minutes before starting and ending the class, even in a real-time environment. This method is efficient than other ML models based on LBPH, Eigen face and Fisher face (Chowdhury et al., 2019). The convolutional neural networks (CNN) based ML method of smart attendance is ideal for online and virtual environments (Srivastava et al., 2019). Advanced ML methods also store multiple face data and can detect multiple faces in real time-consuming time. The principal component analysis (PCA) also takes the follow-up process of keeping the attendance information and redirects to the required email as an attendance sheet. The teachers or instructors should have a detailed summary of students’ attendance in their classes (Sawhney et al., 2019). The smart digital attendance applies CNN with a camera used by the students for more pixel clarity and pleasing light surroundings for detection and recognition (Devi & Narayanan, 2022). The ML supports facial expression detection for monitoring the physical presence of the students. The facial expression analysis gives information to teachers about the number of students present in their classes, not just physical attendance but also collective feedback on to what extent students could cope with the procedures (Ashok et al., 2020).
The ML has a substantial part in assisting smart attendance and an eye on learners’ attention. There is evidence of tracking the student’s attentiveness with ML-based multimodal biometric techniques, including eye gazing and behavioral biometrics involving body movements (Villa et al., 2020). In critical student training such as defense aircraft systems, ML was implemented with wearable sensors for eye tracking, cardiovascular, respiratory, and electrodermal classifiers to predict the accurate levels of learner’s involvement engagement in a training session (Carroll et al., 2020). In air-pilot classes, the machine vision-based ML approaches can be used to determine the students’ attentiveness. The refined identification classifiers such as eye gaze, head poses, and facial expressions can predict the student’s attention level (Goldberg et al., 2019). Any student who attends a class with physical attendance, attentiveness, and zeal, can successfully come up in the upcoming examinations.
Role of ML in Exams (Examination Phase)
The core activities in carrying out the examination system are publishing the examination schedule, question paper generation, authentication of eligible students appearing for the examination, proctoring (supervision) of the examinations, and eliminating malpractices by appropriate steps.
Scheduling examinations
The publication of the examination timetable or schedule is the first step in conducting an examination. Machine learning has played a predominant role in scheduling since 1980. ML with artificial intelligence has proven to be capable of dynamic scheduling of examinations. The ML has already ruled the commercial industries in scheduling, execution, and automation. Now, ML is applied to education, payrolls, generating timetables, and conduction of examinations (Aytug et al., 1994). In universities, students are allowed to study extra credit courses parallelly, which may lead to overlapping in the examination timetable schedule. There will be difficulty in preparing timetables in these overlapping cases manually. The ML has a model that can automatically generate the timetable (Kumar et al., 2020). The ML-based examination schedule system architecture is given in Fig. 7.
The improved automation methods with iterative machine learning and optimization techniques will help conduct a fool-proof online or virtual examination. Neural networks and decision trees have a significant role in this ML-based model’s timetable generation by the iterative development method. ML for the examination process reduces the work burden of the instructors and invigilators to a large extent (Hutter, Kotthoff & Vanschoren, 2019).
Question paper generation
Precaution: The students may use the internet to search exam questions and find answers in google or search engines during the examination. Based on this viewpoint, the questions can be changed with a different verbatim and paraphrased to discourage web searching (Golden & Kohlbeck, 2020). Another method is the generation of randomized multiple-choice questions using the ML method. These questions can form a part of the question paper of the online examination to reduce the repeated questions involving cooperative cheating (Tiong & Lee, 2021).
Authentication of students
The verification of student identity is a pivotal step before the examination to avoid impersonation. ML methods can be implemented for the authentication of student identity. In some complex cases such as twins the multimodal authentication techniques are used for authentication of the individuals. The various existing methods of authentication are listed below in Table 2.
| Paper ID | Year | Authentication | ML methods |
|---|---|---|---|
| Rateria & Agarwal (2018) | 2018 | Offline signature | Hybrid classifier CNN and SVM |
| Gideon et al. (2018) | 2018 | Handwritten signature | Conventional neural networks |
| Bibi, Naz & Rehman (2019) | 2019 | Signature online/offline | ANN, NB, KNN, and SVM |
| Sadikan, Ramli & Fudzee (2019) | 2019 | Keystroke dynamics | ANN, KNN, and hybrid classifiers |
| Ryu, Yeom & Kim (2020) | 2021 | Multi biometric face/keystroke | Eigen Face and SVM |
| Geetha et al. (2021) | 2021 | Face recognition | Eigen Face and SVM |
| Deepika & Sneha (2021) | 2021 | Face detection | Logistic regression |
| Sukmandhani & Sutedja (2019) | 2019 | Face recognition | Self-organized Neural networks |
| Mehta (2019) | 2019 | Face and emotion | Local Binary Pattern Histogram |
| Kamencay et al. (2017) | 2017 | Face recognition | Convolutional Neural Networks |
| Adetoba et al. (2022) | 2020 | Face, video, and password | Logistic regression and SVM |
| Cerna et al. (2018) | 2018 | Face and fingerprint | Multitask Convolutional Neural Networks |
| RaviTeja, Brahmananda & Swasthika Jain (2020) | 2020 | Face and fingerprint | Convolutional neural networks |
| De Marsico, Petrosino & Ricciardi (2016) | 2016 | Iris recognition | ANN and SVM |
| Shelke & Bagal (2017) | 2017 | Iris identification | ANN and SVM |
| Traoré et al. (2017) | 2017 | Face, keyboard, and mouse | Exam shield application |
| Prabu, Lakshmanan & Mohammed (2019) | 2019 | Iris and hand geometry | Neural Networks and Bayes Networks |
| Joshy et al. (2018) | 2018 | Face, one time password, and fingerprint | Histogram of oriented gradients with KNN |
| Ryu et al. (2021) | 2021 | Continuous multimodal biometric | KNN, NB, and Random Forest |
| Labayen et al. (2021) | 2021 | Multimodal | Neural Networks and AI |
DOI: 10.7717/peerjcs.986/table-2
Proctoring (supervision) of examinations
The proctoring process of examinations requires continual monitoring and verification of student identity. The student’s face is initially registered before the commencement of the examination in a database. For continuous validation of a student’s face, an ML-based Convolutional Neural Network (CNN) collects images for face recognition and efficiently verifies the student’s identity (Asep & Bandung, 2019). The ML algorithms capture the face of the student during the examination and match it with the collected images for face recognition to confirm student attendance. This method is automatic and continuous during online proctoring (Ghizlane, Hicham & Reda, 2019).
The ML-based Proctoring software PROCTORU is widely used in online examinations. The ML can record numerous behavioral changes in a student. The head, hand, and eyes movements are tracked and are created as data points with webcam scanning. Suppose the symptom of mismatch pattern is detected, then immediately sent to the knowledge of the admin to keep close watch the abnormal student behavior safeguarding the security and integrity of the examination (Slusky, 2020). The ML with computer vision is ideal for face recognition and gesture detection in image processing. Real-time face recognition is possible by ML in OPENCV proctoring tool. The unsupervised examination is feasible due to ML (Pandey et al., 2020). The PROCTORIO tool with ML is used for live detection of human faces with automation. ML also supports the ID verification process during proctoring examinations. The 360-degree security camera was proposed in online proctoring for more coverage of the surrounding of the student. The sound is recorded, and movements are tracked during an examination. The webcam and security camera can monitor the surrounding of the students with ML algorithms (Turani, Alkhateeb & Alsewari, 2020). This method is advantageous for both live proctoring and recorded proctoring.
The artificial neural networks with ML pave the way for data analytics in the authentication phase of invigilation. The unique ability of ML is to retrieve the required data and label them from the unstructured data is used in recognitions and detections for proctoring tools (Bhardwaj, 2020). In remote online proctoring systems, ML verifies the user and tracks the student’s gazes. When integrated with RESPONDUS, it can lock tabs to prevent cheating by locking down the browser (Vamsi & Ashwin, 2021). Integrated remote supervision with machine learning was proposed in IRS-MLA. This model enables a hybrid learning and examination management system for the student’s maximum benefit, providing a safe environment with remote supervision (Lu, Vivekananda & Shanthini, 2022). The supervision with ML proved to be more efficient and safer during the pandemic times. The online ML-based English hierarchical test was tested against the traditional test system. The comparative analysis showed that ML proved efficient in maintaining fairness and speed (Wang et al., 2021). The evidence suggests ML and ML-based proctoring is influential and trustworthy in all the revisited cases. Emerging AI technologies impact the future of exam management in a far better way (Vincent-Lancrin & Van der Vlies, 2020).
Fraud or cheat detection
The advancements in technology promote remote proctoring and unsupervised methods; there is a susceptibility to cheating and fraud during the examination. Cheating in the online assessment is termed ”Academic Dishonesty”. The cheat detection can be done in-situ or in a-posteriori detection. In live proctoring, the detection should be in-situ, but in many cases, a posteriori detection is done with the help of ANN and SVM (Küppers et al., 2022).
The fraud detection module is designed with ML, and the threshold is set to verify fraud occurrence. The module includes multimodal biometric verification with activity analysis. The detecting module does continuous monitoring, gather data and image to identify any event of fraud in online proctoring (Haytom et al., 2020). In in-situ detection, the head pose detection and the gaze tracking with ML is called VFOA (visual focus of attention). This model has a threshold X for cheat detection. When the VFOA value is more significant than X, it alerts the proctor of possible cheating and makes them more vigilant (Indi et al., 2021).
The articles with ML spoof identifiers are displayed in Table 3.
| Paper ID | Spoof identifier | ML techniques |
|---|---|---|
| Thepade et al. (2020) | Face | Naïve Bayes, SMV, MLP, Decision table, J48, Random Forest & Random Tree |
| Patil et al. (2020) | Face | CNN and HAAR cascade classifier |
| Raj, Tamilselvi & Javith (2022) | Face | 3D face with CNN, LBPH, and HAAR cascade classifier |
| Singh, Joshi & Nandi (2014) | Eye and mouth | Principal component analysis with HAAR cascade |
| Deepika & Kirithigaa (2021) | Retina movement | Logistic regression with data mining |
| Nema (2020) | Blink count | Support Vector Machine |
| Deepika & Philip (2021) | Lip movement | Logistic, Linear, and polynomial regression |
DOI: 10.7717/peerjcs.986/table-3
The assessment of online e-learning is part and parcel of the examination system. Machine learning has transfigured the assessment system. The ML with AI provided a breakthrough in the online grading system with automation and more accurate assessment (Gardner, O’Leary & Yuan, 2021). The DT, NB, and SVM are widely used in evaluation techniques to assess students’ learning outcomes. These algorithms can estimate or predict the student’s expected performance (based on their continuous internal assessment) in advance and are used for validity purposes while doing answer script evaluation (Rana et al., 2021). ML has a significant role in adapting to be integrated with other systems to design innovative assessment models. The Item Response Theory with Natural Language Processing is used to build a student assessment method model (Marigowda et al., 2021). The various ML algorithms used in awarding grades are discussed below:
Privacy is a big question mark since ML applications require personal or private data for student authentication or verification during the examination. These identity data are stored in a location for comparison and verification before examinations. There is a possibility of hacking these data, such as student addresses, that can be exposed or misused. There is a possibility of privacy issues if these data are mishandled (Al-Rubaie & Chang, 2019). The ML algorithms use private data to design the prediction or evaluation models. These training models are prone to attacks. So, there is a need for privacy in two areas: Training set data privacy and model privacy. The adversaries have two types of access to the training models. The white box access is full access to the training set model with knowledge of the data. On the contrary, black-box access is the access to the prediction model with queries and interference.
In ML, the models for prediction and the private data require assured protection of student data (Liu et al., 2020). Another concern is that the student is monitored on video while writing examinations during an assessment. The General Data Protection Regulation (GDPR) has instructed academic institutions to justify the necessity of video recording. The survey indicates that students’ privacy is violated when closely monitored by their devices such as a webcam. Though that is a need of an hour, the student’s privacy should never be broken during the supervision by monitoring their devices wherever they may be (Langenfeld, 2020).
(b) Security
The student data should be ensured private, and privacy always necessitates security against privacy attacks. Privacy and security go hand in hand against threats due to pitfalls in ML. The secure and private data with white box access are attacked less since the attacker has almost all data access on models and private training data. The black box access, the data that are prone to be more secure are targeted for attacks. The ML models should detect such attacks and preserve the model and the privacy of the data (Papernot et al., 2018). The issue with ML is that the assessment models require top-notch security as it contains highly secure data of exam question papers and student data. The disadvantage is that these secure models are cost-heavy. The integrity of these assessment models depends on cost-effectiveness, which may weigh a load on the academic institutions to need more financial support (Langenfeld, 2020). Many attacks or anomalies are induced in the ML system for less security. The attacks are data poisoning, backdoor attacks on the assessment models, interference attacks on sensitive data, and private data stealing. The security methods are inevitable in the ML architecture for security (Xue et al., 2020).
(c) Fairness
In the design of online assessment models based on ML, the neural networks and decision-making algorithms are applied in evaluation. These models should judge and evaluate students’ answers with fairness and integrity. The decision trees are like human judgment, but machine algorithms do not assess the students’ tests with human consideration and sensitivity. The ML decision-making algorithm should be ensured with additional information for more fairness (Mehrabi et al., 2021).
Future solutions of ML
Though ML is vulnerable to attacks, the ML models can detect these breaches and attacks. The ML algorithms can handle authentication, distributed attacks, anomaly, or intrusion detection (Hussain et al., 2020; Liang et al., 2019). With the required additional security features, it can be modified for more security (Kaddoura & Husseiny, 2021; Kaddoura & Al Husseiny, 2021).
(a) Privacy-preserving Machine language (PPML)
When the multiple parties train in an educational assessment of the teaching module, the data can be secured privately with cryptographic approaches. The issues and solutions for the ML are tabulated in Table 4.
| Threats/attacks | Solution | ||
|---|---|---|---|
| Private data in the clear Model inversion Membership De-anonymization | Cryptographic: Homomorphic encryption, Garbled circuits, Secret sharing, Secure processors, | Perturbation approaches: Differential privacy Local differential privacy Dimensionality reduction | |
| Model extraction Feature estimation Membership interference Model Memorization | Cryptographic: Homomorphic Encryption (HE), Cryptonets, Multikey HE | Obfuscation: Differential privacy (DP), Simpler DP, DPfed AVG | Aggregation Federated learning Private aggregation |
DOI: 10.7717/peerjcs.986/table-4
| Attacks | Solution |
|---|---|
| Poisoning attacks | ANTIDOTE, KUAFUDET, AUROR, and Defending SVM |
| Backdoor attacks | Activated clustering method and STRIP |
| Adversarial attacks | Fast gradient sign method and Sec defender |
| Model stealing | ML capsule and PRADA |
| Sensitive data | PATE |
| Misuse attacks | VANET, CANN, and KDD |
| Anomaly attacks | LSTM RNN and RNN IDS |
| Malware attacks | LSTM |
DOI: 10.7717/peerjcs.986/table-5
(b) Security solutions
Standard security attacks are training set poisoning, backdoor attacks, adversarial example attacks, model theft, and recovery of sensitive training data. The available solutions are provided in Table 5.
Conclusions
Machine Learning is instrumental in transforming the current field of education under lockdown scenarios. ML forms a basis for machines to learn from real-time experiences and examples. Its learning capability using statistical techniques paves the way for future education. The benefit to students and the potential of the ML to automate learning and assessments is this article’s focus. The lockdown necessitated Machine Learning to transform the education mode to virtual or online. The complete exam cycle from preparation to feedback is the thrust area for lockdown exam management. ML has proved its significance in smart attendance, identifying academically vulnerable students, personalized learning, and predictive analysis of students’ educational performance in the preparation phase before the examination. Automatic scheduling, authentication, proctoring, and quest bank generation are possible with ML algorithms during the examination phase. The feedback on examinations, fraud, and cheat detection and prevention were highlighted for more understanding. ML algorithms also efficiently evaluate all types of questions, be they a quiz or descriptive. This article ends with a note on the threats and challenges such as privacy, security, and fairness. This article is structured comprehensively with an analysis of machine learning methods for Lockdown Exam Management Systems. The future work is planned to bring together the Deep learning concepts and explore the implication of deep learning in exam management systems.
Additional Information and Declarations
Competing Interests
Jude D. Hemanth is an Academic Editor for PeerJ.
Author Contributions
Sanaa Kaddoura conceived and designed the experiments, performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the paper, and approved the final draft.
Daniela Elena Popescu conceived and designed the experiments, authored or reviewed drafts of the paper, and approved the final draft.
Jude D. Hemanth analyzed the data, prepared figures and/or tables, and approved the final draft.
Data Availability
The following information was supplied regarding data availability:
This is a literature review.
Funding
This work was funded by the Zayed University Research Incentive Fund (RIF) grant (grant number R20128). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Adetoba BT, Awodele O, Kuyoro SO, Nwaocha OV. 2022. An improved authentication and monitoring system for E-learning examination using supervised machine learning algorithms. International Journal of Scientific & Engineering Research 11(3):236-242
Adnan M, Habib A, Ashraf J, Mussadiq S, Raza AA, Abid M, Bashir M, Khan SU. 2021. Predicting at-risk students at different percentages of course length for early intervention using machine learning models. IEEE Access 9:7519-7539
Agarwal S, Punn NS, Sonbhadra SK, Tanveer M, Nagabhushan P, Pandian KKS, Saxena P. 2021. Unleashing the power of disruptive and emerging technologies amid COVID-19: a detailed review. preprint
Ahadi A, Lister R, Haapala H, Vihavainen A. 2015. Exploring machine learning methods to automatically identify students in need of assistance. In: Proceedings of the eleventh annual international conference on international computing education research.
Akour I, Alshurideh M, Al Kurdi B, Al Ali A, Salloum S. 2021. Using machine learning algorithms to predict people’s intention to use mobile learning platforms during the COVID-19 pandemic: machine learning approach. JMIR Medical Education 7(1):e24032
Al-Rubaie M, Chang JM. 2019. Privacy-preserving machine learning: threats and solutions. IEEE Security & Privacy 17(2):49-58
Al-Shabandar R, Hussain AJ, Liatsis P, Keight R. 2019. Detecting at-risk students with early interventions using machine learning techniques. IEEE Access 7:149464-149478
Albreiki B, Zaki N, Alashwal H. 2021. A systematic literature review of student’ performance prediction using machine learning techniques. Education Sciences 11(9):552
Alnassar F, Blackwell T, Homayounvala E, Yee-king M. 2021. How well a student performed? A machine learning approach to classify students’ performance on virtual learning environment. In: 2021 2nd international conference on intelligent engineering and management (ICIEM).
Alshabandar R, Hussain A, Keight R, Khan W. 2020. Students performance prediction in online courses using machine learning algorithms. In: 2021 2nd international conference on intelligent engineering and management (ICIEM).
Anzén E. 2022. The viability of machine learning models based on levenstein distance and cosine similarity for plagiarism detection in digital exams. (accessed31 March 2022 )
Asep HSG, Bandung Y. 2019. A design of continuous user verification for online exam proctoring on M-learning. In: 2019 International Conference on Electrical Engineering and Informatics (ICEEI).
Ashok KS, Ragul RN, Gokula KD, Praveen KS. 2020. Student monitoring system using machine learning. International Journal of Innovative Technology and Exploring Engineering 9(4):1475-1479
August SE, Tsaima A. 2021. Artificial intelligence and machine learning: an instructor’s exoskeleton in the future of education. In: Innovative learning environments in STEM higher education. 79-105
Aytug H, Bhattacharyya S, Koehler GJ, Snowdon JL. 1994. A review of machine learning in scheduling. IEEE Transactions on Engineering Management 41(2):165-171
Baashar Y, Alkawsi GA, Mustafa AS, Alkahtani AA, Alsariera YA, Ali AQ, Hashim W, Tiong SK. 2022. Toward predicting student’s academic performance using Artificial Neural Networks (ANNs) Applied Sciences 12(13):1289
Balaji P, Alelyani S, Qahmash A, Mohana M. 2021. Contributions of machine learning models towards student academic performance prediction: a systematic review. Applied Sciences 11(21):10007
Bashir MF, Arshad H, Javed AR, Kryvinska N, Band SS. 2021. Subjective answers evaluation using machine learning and natural language processing. IEEE Access 9:158972-158983
Bashitialshaaer R, Alhendawi M, Lassoued Z. 2021. Obstacle comparisons to achieving distance learning and applying electronic exams during COVID-19 pandemic. Symmetry 13(1):99
Berens J, Schneider K, Görtz S, Oster S, Burghoff J. 2018. Early detection of students at risk—predicting student dropouts using administrative student data and machine learning methods. SSRN Electronic Journal 11:1-41
Bhardwaj R. 2020. Online exam management system using deep learning. International Journal of Creative Research Thoughts 8:4185
Bibi K, Naz S, Rehman A. 2019. Biometric signature authentication using machine learning techniques: current trends, challenges and opportunities. Multimedia Tools and Applications 79:289-340
Carroll M, Ruble M, Dranias M, Rebensky S, Chaparro M, Chiang J, Winslow B. 2020. Automatic detection of learner engagement using machine learning and wearable sensors. Journal of Behavioral and Brain Science 10(03):165-178
Çınar A, Ince E, Gezer M, Yılmaz Ö. 2020. Machine learning algorithm for grading open-ended physics questions in Turkish. In: Education and information technologies. New York: Springer.
Cerna P, Charlemaine M, Gebreziagbher H, Mengstie M. 2018. Machine learning biometric attendance system using fingerprint fuzzy vault scheme algorithm and multi-task convolution neural network face recognition algorithm. International Journal of Computer Applications 179(48):1-6
Chauhan R, Saharan R, Singh S, Sharma P. 2020. Automated content grading using machine learning. preprint
Chen C, Yang SJ, Weng J, Ogata H, Su CY. 2021. Predicting at-risk university students based on their e-book reading behaviours by using machine learning classifiers. Australian Journal of Educational Technology 37(4):130-144
Chowdhury AE, Chowdhury OK, Samrat A, Rahman Z, Ahmed T. 2019. HFDCM: a low-cost machine learning based class attendance monitoring system.
Chui KT, Fung DCL, Lytras MD, Lam TM. 2020. Predicting at-risk university students in a virtual learning environment via a machine learning algorithm. Computers in Human Behavior 107:105584
De Marsico M, Petrosino A, Ricciardi S. 2016. Iris recognition through machine learning techniques: a survey. Pattern Recognition Letters 82:106-115
Deepika M, Kirithigaa K. 2021. Eye retina movement detection on online examination along machine learning. JETIR 8:1-4
Deepika M, Sneha H. 2021. Journal of Emerging Technologies and Innovative Research 8:1-5
Deepika S, Philip P. 2021. Lip movement detection on online exam along with machine learning. Journal of Emerging Technologies and Innovated Research 8:1-4
Devi VA, Narayanan SS. 2022. A smart digital attendance monitoring system for academic institution using machine learning techniques. Webology 19(1):3109-3118
Dhilipan J, Vijayalakshmi N, Suriya S, Christopher A. 2021. Prediction of students performance using machine learning. IOP Conference Series: Materials Science and Engineering 1055(1):012122
Embarak O. 2021. Towards an adaptive education through a machine learning recommendation system. In: 2021 international conference on artificial intelligence in information and communication (ICAIIC). Piscataway. IEEE. 187-192
Fagbola TM, Adeyanju I, Olaniyan O, Esan A, Omodunbi B, Oloyede A, Egbetola F. 2019. Development of mobile-interfaced machine learning-based predictive models for improving students performance in programming courses. preprint
Falkner NJG, Falkner KE. 2012. A fast measure for identifying at-risk students in computer science. In: Proceedings of the ninth annual international conference on international computing education research—ICER ’12.
Galhardi LB, Brancher JD. 2018. Machine learning approach for automatic short answer grading: a systematic review. Advances in Artificial Intelligence—IBERAMIA 2018:380-391
Gardner J, O’Leary M, Yuan L. 2021. Artificial intelligence in educational assessment: breakthrough? Or buncombe and ballyhoo? Journal of Computer Assisted Learning 37(5):1207-1216
Geetha M, Latha RS, Nivetha SK, Hariprasath S, Gowtham S, Deepak CS. 2021. Design of face detection and recognition system to monitor students during online examinations using Machine Learning algorithms. In: 2021 international conference on computer communication and informatics (ICCCI). Piscataway. IEEE. 1-4
Ghanta H. 2019. Automated essay evaluation using natural language processing and machine learning. thesis
Ghizlane M, Hicham B, Reda FH. 2019. A new model of automatic and continuous online exam monitoring. In: 2019 International Conference on Systems of Collaboration Big Data, Internet of Things & Security (SysCoBIoTS).
Gideon SJ, Kandulna A, Kujur AA, Diana A, Raimond K. 2018. Handwritten signature forgery detection using convolutional neural networks. Procedia Computer Science 143:978-987
Goldberg P, Sümer Ö, Stürmer K, Wagner W, Göllner R, Gerjets P, Kasneci E, Trautwein U. 2019. Attentive or not? Toward a machine learning approach to assessing students’ visible engagement in classroom instruction. Educational Psychology Review 33(1):27-49
Golden J, Kohlbeck M. 2020. Addressing cheating when using test bank questions in online classes. Journal of Accounting Education 52:100671
Haytom M, Rosenberger C, Charrier C, Zhu C, Regnier C. 2020. Identity verification and fraud detection during online exams with a privacy compliant biometric system. In: Proceedings of the 17th international joint conference on e-business and telecommunications.
Ho IMK, Cheong KY, Weldon A. 2021. Predicting student satisfaction of emergency remote learning in higher education during COVID-19 using machine learning techniques. PLOS ONE 16(4):e0249423
Hussain F, Hussain R, Hassan SA, Hossain E. 2020. Machine learning in IoT security: current solutions and future challenges. IEEE Communications Surveys Tutorials 22(3):1686-1721
Hussain M, Zhu W, Zhang W, Abidi SMR, Ali S. 2018. Using machine learning to predict student difficulties from learning session data. Artificial Intelligence Review 52(1):381-407
Hussain S, Khan MQ. 2021. Student-performulator: predicting students’ academic performance at secondary and intermediate level using machine learning. In: Annals of data science. New York: Springer. Epub ahead of print Apr 21 2021
Hutter F, Kotthoff L, Vanschoren J. 2019. Automated machine learning. New York: Springer International Publishing.
Hylton K, Levy Y, Dringus LP. 2016. Utilizing webcam-based proctoring to deter misconduct in online exams. Computers & Education 92-93:53-63
Indi CS, Pritham V, Acharya V, Prakasha K. 2021. Detection of malpractice in e-exams by head pose and gaze estimation. International Journal of Emerging Technologies in Learning (iJET) 16(08):47-60
Jain A, Solanki S. 2019. An efficient approach for multiclass student performance prediction based upon machine learning. In: 2019 International Conference on Communication and Electronics Systems (ICCES).
Joshy N, Kumar G, Mukhilan P, Prasad M, Ramasamy T, Student H. 2018. Multi-factor authentication scheme for online examination. International Journal of Pure and Applied Mathematics 119(15):1705-1712
Kabudi T, Pappas I, Olsen DH. 2021. AI-enabled adaptive learning systems: a systematic mapping of the literature. Computers and Education: Artificial Intelligence 2:100017
Kaddoura S, Al Husseiny F. 2021. An approach to reinforce active learning in higher education for IT students. Global Journal of Engineering Education 23(1):43-48
Kaddoura S, Husseiny FAl. 2021. On-line elarning on information security based on critical thinking andragogy. World Transactions on Engineering and Technology Education 19(2):157-162
Kamalov F, Sulieman H, Santandreu Calonge D. 2021. Machine learning based approach to exam cheating detection. PLOS ONE 16(8):e0254340
Kamencay P, Benco M, Mizdos T, Radil R. 2017. A new method for face recognition using convolutional neural network. Advances in Electrical and Electronic Engineering 15(4)
Karalar H, Kapucu C, Gürüler H. 2021. Predicting students at risk of academic failure using ensemble model during pandemic in a distance learning system. International Journal of Educational Technology in Higher Education 18(1):63
Kondo N, Okubo M, Hatanaka T. 2017. Early detection of at-risk students using machine learning based on LMS log data. In: 2017 6th IIAI international congress on advanced applied informatics (IIAI-AAI). Piscataway. IEEE. 198-201
Kotsiantis S, Pierrakeas C, Pintelas P. 2004. Predicting students’ performance in distance learning using machine learning techniques. Applied Artificial Intelligence 18(5):411-426
Koutcheme C, Sarsa S, Hellas A, Haaranen L, Leinonen J. 2022. Methodological considerations for predicting at-risk students. In: Australasian Computing Education Conference.
Koutina M, Kermanidis KL. 2011. Predicting postgraduate students’ performance using machine learning techniques. In: IFIP Advances in Information and Communication Technology. 159-168
Kumar P, Sanakar S, Kumar P, Usman SM. 2020. Automated timetable generator using machine learning.
Küppers B, Opgen-Rhein J, Eifert T, Schroeder U. 2022. Cheating detection: identifying fraud in digital exams. (accessed31 March 2022 )
Labayen M, Vea R, Florez J, Aginako N, Sierra B. 2021. Online student authentication and proctoring system based on multimodal biometrics technology. IEEE Access 9:72398-72411
Langenfeld T. 2020. Internet-based proctored assessment: security and fairness issues. Educational Measurement: Issues and Practice 39:24-27
Latif S, XianWen F, Wang L. 2021. Intelligent decision support system approach for predicting the performance of students based on three-level machine learning technique. Journal of Intelligent Systems 30(1):739-749
Lavoué E, Drachsler H, Verbert K, Broisin J, Pérez-Sanagustín M. 2017. Data drive approaches in digital education. In: 12th european conference on technology enhanced learning, EC-TEL 2017. Berlin, Heidelberg. Springer International Publishing.
Li H, Ding W, Yang S, Liu Z. 2020. Identifying at-risk K-12 students in multimodal online environments: a machine learning approach. preprint
Li S, Liu T. 2021. Performance prediction for higher education students using deep learning. Complexity 2021:1-10
Liang F, Hatcher WG, Liao W, Gao W, Yu W. 2019. Machine learning for security and the Internet of Things: the good, the bad, and the ugly. IEEE Access 7:158126-158147
Linardatos P, Papastefanopoulos V, Kotsiantis S. 2020. Explainable AI: a review of machine learning interpretability methods. Entropy 23(1):18
Lincke A, Jansen M, Milrad M, Berge E. 2021. The performance of some machine learning approaches and a rich context model in student answer prediction. Research and Practice in Technology Enhanced Learning 16(1):1-18
Liu B, Ding M, Shaham S, Rahayu W, Farokhi F, Lin Z. 2020. When machine learning meets privacy: a survey and outlook. preprint
Louridas P, Ebert C. 2016. Machine learning. IEEE Software 33(5):110-115
Lu W, Vivekananda GN, Shanthini A. 2022. Supervision system of English online teaching based on machine learning. In: Progress in artificial intelligence. New York: Springer.
Lv H, Tang H. 2011. Machine learning methods and their application research. In: 2011 2nd international symposium on intelligence information processing and trusted computing. Piscataway. IEEE. 108-110
Marigowda CK, Ranjit A, Ashwin Prasad A, Karanth B, Sharma S. 2021. A comprehensive examination assessment model using machine learning. International Journal of Engineering Research & Technology 10(1):1-10
Marijan D, Gotlieb A, Kumar Ahuja M. 2019. Challenges of testing machine learning based systems. In: 2019 IEEE international conference on artificial intelligence testing (AITest). Piscataway. IEEE. 101-102
Marwaha A, Singla A. 2019. A study of factors to predict at-risk students based on machine learning techniques. In: Advances in intelligent systems and computing. New York: Springer. 133-141
Mehrabi N, Morstatter F, Saxena N, Lerman K, Galstyan A. 2021. A survey on bias and fairness in machine learning. ACM Computing Surveys 54(6):1-35
Mehta R. 2019. Real-time student surveillance system using machine learning and computer vision. International Journal of Advanced Research in Computer Science 10(4):29-33
Mirata V, Hirt F, Bergamin P, Van der Westhuizen C. 2020. Challenges and contexts in establishing adaptive learning in higher education: findings from a Delphi study. International Journal of Educational Technology in Higher Education 17(1):32
Nema A. 2020. Ameliorated anti-spoofing application for PCs with users’ liveness detection using blink count. In: 2020 international conference on computational performance evaluation (ComPE). Piscataway. IEEE. 311-315
Oboko RO, Maina EM, Waiganjo PW, Omwenga EI, Wario RD. 2016. Designing adaptive learning support through machine learning techniques.
Oke I, Kulkarni P, Lolap T, Haribhakta Y. 2016. Adaptive learning system for analysis of performance and learning behavior of school children.
Pallathadka H, Wenda A, Ramirez-Asís E, Asís-López M, Flores-Albornoz J, Phasinam K. 2021. Classification and prediction of student performance data using various machine learning algorithms. In: Materials today: proceedings.
Pandey AK, Kumar S, Rajendran B, Bindhumadhava BS. 2020. e-Parakh: unsupervised online examination system. In: 2020 IEEE region 10 conference (TENCON). Piscataway. IEEE. 667-671
Papernot N, McDaniel P, Sinha A, Wellman MP. 2018. SoK: Security and Privacy in Machine Learning. In: 2018 IEEE European symposium on security and privacy (EuroS & P). Piscataway. IEEE.
Patil P, Desale S, Kotkar V, Alkunte R, Gutte V. 2020. An approach for face liveness detection and face recognition using machine learning algorithm. The International Journal of Creative Research Thoughts 8(6):2320-2882
Peng H, Ma S, Spector JM. 2019. Personalized adaptive learning: an emerging pedagogical approach enabled by a smart learning environment. Smart Learning Environments 6(1):9
Phalke A, Bamnote G, Ahmad SW. 2021. Evaluating descriptive answer assessment system with machine learning. International Journal of Creative Research Thoughts 9(7):2320-2882
Prabu S, Lakshmanan M, Mohammed VN. 2019. A multimodal authentication for biometric recognition system using intelligent hybrid fusion techniques. Journal of Medical Systems 43(8):249
Priya S, Ankit T, Divyansh D. 2021. Student performance prediction using machine learning. In: Advances in parallel computing technologies and applications. Amsterdam, Netherlands: IOS Press.
Qiu F, Zhang G, Sheng X, Jiang L, Zhu L, Xiang Q, Jiang B, Chen P. 2022. Predicting students’ performance in e-learning using learning process and behaviour data. Scientific Reports 12(1):453
Raj S, Tamilselvi K, Javith S. 2022. Face identification and liveness detection using cnn for automated attendance system. International Journal of Advance
Rana P, Raj Gupta L, Dubey MK, Kumar G. 2021. Review on evaluation techniques for better student learning outcomes using machine learning. In: 2021 2nd international conference on intelligent engineering and management (ICIEM). Piscataway. IEEE. 86-90
Rateria A, Agarwal S. 2018. Offline signature verification through machine learning. In: 2018 5th IEEE Uttar Pradesh section international conference on electrical, electronics and computer engineering (UPCON). Piscataway. IEEE. 1-7
RaviTeja K, Brahmananda SH, Swasthika Jain T. 2020. Multi-model biometric authentication system for smart attendance. International Journal of Recent Technology and Engineering 8(6):4284-4287
Rosy Salomi Victoria D, Viola Grace Vinitha P, Sathya R. 2020. Intelligent short answer assessment using machine learning. International Journal of Engineering and Advanced Technology 9(4):1111-1116
Ruipérez-Valiente JA, Muñoz Merino PJ, Alexandron G, Pritchard DE. 2019. Using machine learning to detect multiple-account cheating and analyze the influence of student and problem features. IEEE Transactions on Learning Technologies 12(1):112-122
Ryu R, Yeom S, Kim S-H. 2020. Continuous multibiometric authentication for online exam with machine learning.
Ryu R, Yeom S, Kim S-H, Herbert D. 2021. Continuous multimodal biometric authentication schemes: a systematic review. IEEE Access 9:34541-34557
Sadikan SFN, Ramli AA, Fudzee MF. 2019. A survey paper on keystroke dynamics authentication for current applications. In: Advances in electrical and electronic engineering: from theory to applications (SERIES 2): Proceedings of the International Conference of Electrical and Electronic Engineering (ICon3E 2019).
Sandra L, Lumbangaol F, Matsuo T. 2021. Machine learning algorithm to predict student’s performance: a systematic literature review. TEM Journal 10:1919-1927
Sarker IH. 2021. Machine learning: algorithms, real-world applications and research directions. SN Computer Science 2(3):160
Sawhney S, Kacker K, Jain S, Singh SN, Garg R. 2019. Real-time smart attendance system using face recognition techniques. In: 2019 9th international conference on cloud computing, data science & engineering (Confluence). Piscataway. IEEE. 522-525
Sekeroglu B, Dimililer K, Tuncal K. 2019. Student performance prediction and classification using machine learning algorithms. In: Proceedings of the 2019 8th international conference on educational and information technology—ICEIT 2019.
Shashi S, Sunil KP, Kumkum G. 2020. Machine learning for predictions in academics. International Journal of Recent Technology and Engineering 8:1-4
Shelke R, Bagal SB. 2017. Iris recognition system: a novel approach for biometric authentication. Piscataway. IEEE.
Singh AK, Joshi P, Nandi GC. 2014. Face recognition with liveness detection using eye and mouth movement. In: 2014 international conference on signal propagation and computer technology (ICSPCT 2014). Piscataway. IEEE. 592-597
Slusky L. 2020. Cybersecurity of online proctoring systems. Journal of International Technology and Information Management 29:56-83
Soobramoney R, Singh A. 2019. Identifying students at-risk with an ensemble of machine learning algorithms. In: 2017 international conference on computing, communication, control and automation (ICCUBEA). Piscataway. IEEE. 1-5
Srivastava T, Vaish V, Sharma P, Khanna P. 2019. Implementing machine learning for face recognition based attendance monitoring system. In: 2019 6th international conference on computing for sustainable global development (INDIACom). Piscataway. IEEE. 1254-1259
Sukmandhani AA, Sutedja I. 2019. Face recognition method for online exams. In: 2019 International Conference on Information Management and Technology (ICIMTech).
Sweeney M, Lester J, Rangwala H, Johri A. 2016. Next-term student performance prediction: a recommender systems approach. Journal of Educational Data Mining 8(1)
Tarik A, Aissa H, Yousef F. 2021. Artificial intelligence and machine learning to predict student performance during the COVID-19. Procedia Computer Science 184:835-840
Thepade S, Jagdale P, Bhingurde A, Erandole S. 2020. Novel face liveness detection using fusion of features and machine learning classifiers. In: 2020 IEEE international conference on informatics, IoT, and enabling technologies (ICIoT). Piscataway. IEEE. 141-145
Tiong LC, Lee HJ. 2021. E-cheating prevention measures: detection of cheating at online examinations using deep learning approach—a case study. preprint
Traoré I, Nakkabi Y, Saad S, Sayed B, Ardigo JD, De Faria Quinan PM. 2017. Ensuring online exam integrity through continuous biometric authentication. In: Information security practices. New York: Springer. 73-81
Turani AA, Alkhateeb JH, Alsewari AA. 2020. Students online exam proctoring: a case study using 360 degree security cameras. In: 2020 emerging technology in computing, communication and electronics (ETCCE).
Vamsi M, Ashwin S, Kajendran Student. 2021. Remote online proctoring system. International Journal of Creative Research Thoughts 9:2320-2882
Villa M, Gofman M, Mitra S, Almadan A, Krishnan A, Rattani A. 2020. A survey of biometric and machine learning methods for tracking students’ attention and engagement. In: 2020 19th IEEE international conference on machine learning and applications (ICMLA). Piscataway. IEEE.
Vincent-Lancrin S, Van der Vlies R. 2020. Trustworthy artificial intelligence (AI) in education: promises and challenges.
Wang X, Zhang D, Asthana A, Asthana S, Khanna S, Verma CS. 2021. Design of English hierarchical online test system based on machine learning. Journal of Intelligent Systems 30:793-807
Webb ME, Fluck A, Magenheim J, Malyn-Smith J, Waters J, Deschênes M, Zagami J. 2020. Machine learning for human learners: opportunities, issues, tensions and threats. Educational Technology Research and Development 69:2109-2130
Wei Y, Qi Y, Ma Q, Liu Z, Shen C, Fang C. 2020. Fraud detection by machine learning. In: 2020 2nd international conference on machine learning, big data and business intelligence (MLBDBI). Piscataway. IEEE. 101-115
Xiao W, Ji P, Hu J. 2021. A survey on educational data mining methods used for predicting students’ performance. Engineering Reports 4:e12482
Xu J, Moon KH, Van der Schaar M. 2017. A machine learning approach for tracking and predicting student performance in degree programs. IEEE Journal of Selected Topics in Signal Processing 11(5):742-753
Xue M, Yuan C, Wu H, Zhang Y, Liu W. 2020. Machine learning security: threats, countermeasures, and evaluations. IEEE Access 8:74720-74742
Yanes N, Mostafa AM, Ezz M, Almuayqil SN. 2020. A machine learning-based recommender system for improving students learning experiences. IEEE Access 8:201218-201235
Zea LDF, Reina YFP, Molano JIR. 2019. Machine learning for the identification of students at risk of academic desertion. In: Communications in computer and information science. New York: Springer. 462-473
Sanaa Kaddoura1, Daniela Elena Popescu2, Jude D. Hemanth3
1 College of Technological Innovation, Zayed University, Abu Dhabi, United Arab Emirates
2 Faculty of Electrical Engineering and Information Technology, University of Oradea, Oradea, Romania
3 Electronics and Communication Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India
© 2022 Kaddoura et al. This is an open access article distributed under the terms of the Creative Commons Attribution License: https://creativecommons.org/licenses/by/4.0/ (the “License”), which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. For attribution, the original author(s), title, publication source (PeerJ Computer Science) and either DOI or URL of the article must be cited. Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.