Available online at www.sciencedirect.

com
Available online at www.sciencedirect.com
Available
Available online
online at www.sciencedirect.com
at www.sciencedirect.com

ScienceDirect
Procedia Computer Science 00 (2024) 000–000
Procedia Computer Science 00 (2024) 000–000 www.elsevier.com/locate/procedia
Procedia Computer
Procedia Science
Computer 235 (2024)
Science 1399–1408
00 (2024) 000–000 www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia
International Conference on Machine Learning and Data Engineering (ICMLDE 2023)
International Conference on Machine Learning and Data Engineering (ICMLDE 2023)
International
International Conference
Conference on
on Machine
Machine Learning
Learning and
and Data
Data Engineering
Engineering (ICMLDE
(ICMLDE 2023)
2023)
Enhancing Academic Integrity in Online Assessments: Introducing
Enhancing
Enhancing Academic
Academic Integrity
Integrity in
in Online
Online Assessments:
Assessments: Introducing
Introducing
an Effective Online Exam Proctoring Model using YOLO
an Effective
an Effective Online
Online Exam Proctoring Model using YOLO
Tripty Singha,
, RekhaExam
R NairbProctoring ModelDuraisamy
, Tina Babub , Prakash
a,
using YOLO
b
c
Tripty Singh , Rekha
Sciencea,and
R Nair , Tina
bSchool Babubb,Bengaluru,
Prakash Duraisamy c
c
a Department
Tripty
of Singh
Computer , Rekha R
Engineering, Nair
Amrita , Tina Babu
of Computing, , Prakash Duraisamy
Amrita Vishwa Vidyapeetham, India
a Department of Computer Science and Engineering, Amrita School of Computing, Bengaluru, Amrita Vishwa Vidyapeetham, India
b
Department of Computer Science and Engineering, Alliance University, Bengaluru, India
a Departmentc of Computer
b Science and Engineering, Amrita School of Computing, Bengaluru, Amrita Vishwa Vidyapeetham, India
Department
Department of Computer
of Computer ScienceScience and Engineering,
and Engineering, Alliance
University University, Bengaluru,
of Wisconsin-Green India USA
Bay, Wisconsin,
b Department of Computer Science and Engineering, Alliance University, Bengaluru, India
c Department of Computer Science and Engineering, University of Wisconsin-Green Bay, Wisconsin, USA
c Department of Computer Science and Engineering, University of Wisconsin-Green Bay, Wisconsin, USA

Abstract
Abstract
The issue of cheating during exams has become a significant concern, gaining widespread attention from the public and media.
Abstract
The
This issue
issue ofhascheating
been made during exams
worse has transition
by the become aofsignificant
educational concern, gaining
institutions to widespread attention
an online format, fromitthe
making publictoand
simpler media.
engage in
The issue
This
academic of
hascheating
issuedishonesty.
been made during
To exams
worse
overcome hasproblem,
bythis
the become developed
transition aofsignificant anconcern,
educational gaining
institutions
innovative to an
solutionwidespread
online asattention
knownformat, from
making
the Online itthe
Exam public
simpler toand
Proctoring media.
engage
Modelin
This issue
academic
using has been
You dishonesty.
Only Lookmade
To
Once worse
overcome
(YOLO) bythis
the transition
problem,
Model. of educational
Withdeveloped
the use of an institutions
innovative
cutting-edge to an online
solution
computer vision format,
known making
as the Online
algorithms, it simpler
theExam to engage
Proctoring
proposed model Model
makesin
academic
using
use datadishonesty.
ofYou Only
fromLook To
Once
cameras overcome
(YOLO)
and this problem,
Model.
microphones With developed
the use of an
to automatically innovative
cutting-edge
spot solution
computer
and prevent known as the online
visionduring
cheating Online
algorithms, Exam
theexams. Proctoring
proposed
Facemodel Model
makes
orientation,
using
use
face of You
dataOnly
spoofing from Look
detection, Once
cameras (YOLO)
and
mouth Model.
microphones
tracking, With the usemobile
to analysis,
audio automatically of cutting-edge
spot computer
and recognition,
device prevent vision
cheating algorithms,
during
and person onlinethe
counting proposed
exams.
are Facemodel
among makes
orientation,
the primary
use of
face
elements data
spoofing from
built intocameras
detection, and Object
mouth
the model. microphones
tracking, to analysis,
audio
detection automatically
algorithms, in spot
mobile and recognition,
device
conjunction prevent
with cheating during
and person
pre-trained YOLO online
counting exams.
weights, Face the
areempower
among orientation,
theprimary
model
face
to spoofing
elements
detect built
mobile detection,
intodevices mouth
the model.
and counttracking,
Object the audio analysis,
detection
number algorithms, mobile
of individuals device in
in conjunction
present recognition,
withexam
the and person
pre-trained YOLO
environment. counting
weights,
The areempower
among
automatic the
theprimary
model
termination of
elements
to detect
exams uponbuilt
mobile into
detection the
devices model.
of mobile Object
and count detection
the number
devices, algorithms,
multipleof individuals
persons, in conjunction
or no present
persons in with
forthe pre-trained
10 exam
seconds YOLO
environment. weights,
Thedeterrent
acts as a strong empower
automatic the
termination
against model
of
cheating.
to detect
exams
The uponmobile
suggested devices
detection
methodology andcreated
of mobile count athe
devices, numberandof
multiple
reliable individuals
persons,
effective no present
ortool persons
for onlinein the exam environment.
forexaminations,
10 seconds acts Thethe
as a strong
guaranteeing automatic
deterrent termination
against
academic of
cheating.
integrity and
exams
The
enabling upon
suggested detection
impartial of mobile
methodology
assessment ofdevices,
created multiple
the askills
reliable
and and persons,
effectiveof
knowledge ortool
no for
persons
learners. Byfor
online 10 seconds acts
examinations,
implementing as a strongeducational
guaranteeing
this solution, deterrent
the against
academic cheating.
integrity
institutions and
can
The suggested
enabling
foster impartial
a secure methodology
assessment
online created
of the askills
exam environment, reliable
and and effective
knowledge
promoting oftool
a culture offor
learners. online examinations,
By implementing
honesty and guaranteeing
this
trust between solution, the educators.
academic
students educational
and integrity
institutions and
can
As navigate
enabling
foster a impartial
secure assessment
online exam of the
environment,skills and knowledge
promoting a of
culture learners.
of By
honesty implementing
and trust this
between solution,
students
the evolving environment of education, the model stands as a reliable ally in upholding the credibility and authenticity of online educational
and institutions
educators. As can
navigate
foster
the a secure
evolving
learning online exam
environment
experiences. of environment,
education, thepromoting
model stands a culture of honesty
as a reliable ally and trust between
in upholding students and
the credibility andeducators.
authenticityAsofnavigate
online
the evolving
learning environment of education, the model stands as a reliable ally in upholding the credibility and authenticity of online
experiences.
learning experiences.
© 2024 The Authors. Published by Elsevier B.V.
©
© 2024
This The
is an
2024 TheopenAuthors. Published
accessPublished
Authors. article under by ELSEVIER
by Elsevier B.V.B.V. license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
the CC BY-NC-ND
© 2024
This is The
an
Peer-review openAuthors.
underaccess Published
article
responsibilityunderby
of Elsevier
the CC B.V. committee
BY-NC-ND
scientific license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
of the International Conference on Machine Learning and Data Engi-
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
This is
neering. an
Peer-review open
Peer-reviewunder access article
underresponsibility under
responsibilityofof the
thetheCC BY-NC-ND
scientific committee
scientific license
committee of the(http://creativecommons.org/licenses/by-nc-nd/4.0/)
of International
the Conference
International on Machine
Conference Learning
on Machine and Data
Learning andEngi-
Data
Peer-review
Engineeringunder responsibility of the scientific committee of the International Conference on Machine Learning and Data Engi-
neering.
Keywords:
neering. You Only Look Once; Online Exam Proctoring; Mediapipe; Antispoofing; Pyaudio
Keywords: You Only Look Once; Online Exam Proctoring; Mediapipe; Antispoofing; Pyaudio
Keywords: You Only Look Once; Online Exam Proctoring; Mediapipe; Antispoofing; Pyaudio

1. Introduction
1. Introduction
1. Introduction
Online exam proctoring is a technology-driven solutions that enables educational institutions and organizations to
Onlineand
monitor exam proctoring
secure is a technology-driven
online exams[33]. solutions
With the increase in that enables
growth educational
of online institutions
education and work
and remote organizations to
this online
Online
monitor andexam proctoring
secure online is a technology-driven
exams[33]. With the solutions
increase in that enables
growth of educational
online institutions
education and andwork
remote organizations
this to
online
exam proctoring became popular to minimize the cheating during exam[34].
monitor
exam and secure
proctoring online
became exams[33].
popular With the
to minimize theincrease
cheatinginduring
growth of online education and remote work this online
exam[34].
exam proctoring became popular to minimize the cheating during exam[34].
E-mail address: [email protected]
E-mail address: [email protected]
E-mail address:
1877-0509 [email protected]
© 2024 The Authors. Published by Elsevier B.V.
1877-0509
This © 2024
is an open
1877-0509 © 2024 Thearticle
access
The Authors. Published
under by Elsevier
ELSEVIER
the CC BY-NC-ND
by B.V.(http://creativecommons.org/licenses/by-nc-nd/4.0/)
license
B.V.
Thisisisananopen
1877-0509
Peer-review
This open
©underaccess
2024 article
Thearticle
Authors.
responsibility
access under
of
under theBY-NC-ND
Published
CC CCElsevier
by
the scientific BY-NC-ND
B.V.of license
committee
license the (https://creativecommons.org/licenses/by-nc-nd/4.0)
International Conference on Machine Learning and Data Engineering.
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Peer-review
This is an open
Peer-review under
access
under responsibility
article under
responsibility of BY-NC-ND
the committee
the scientific
of the CC scientific committee
license
of of the International Conference on Data
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
the International Conference on Machine Learning and Machine Learning and
Engineering.
Data Engineering
Peer-review under responsibility of the scientific committee of the International Conference on Machine Learning and Data Engineering.
10.1016/j.procs.2024.04.131
1400 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408
2 Tripty et al. / Procedia Computer Science 00 (2024) 000–000

The online exam proctoring project use the specialized software and tools to monitor the students behavior during
the exam[31]. This system able to track the students activity during exam. It can also use artificial intelligence algo-
rithms to find the suspicious behavior, such as looking away from the screen, opening mouth or not, face orientation
ad having someone else in room. This system can detect any suspicious behavior that violates the exam rules[1]. The
online exam proctoring project has several benefits, including increased accessibility, reduced administrative work-
load, and enhanced credibility of online learning[32]. It provides a fair and equal opportunity for students to check
their knowledge without the risk of cheating[19]. However, it is very important to ensure that the system is user-
friendly and it is also accessible to all students, including for the ones with disabilities. It is also important to balance
the benefits of the proctoring system with the concerns about privacy and security, as the system involves the col-
lection and analysis of sensitive data. Many machine learning, Deep learning algorithms are utilized for online exam
proctoring[14, 16, 26, 30, 5, 23, 6, 13, 29].
Academic dishonesty encompasses a wide range of actions and behaviors by students that involve deception or
unethical practices aimed at taking credit for academic work that they did not genuinely create or produce. This
includes actions such as plagiarism, cheating on exams, fabricating data or research findings, submitting someone
else’s work as their own, or engaging in any form of academic fraud or deceitful conduct with the intention of gaining
an unfair academic advantage[20]. Many IOT related devises are utilized for exam proctoring system [27].
Now a days online exam proctoring topic is an active research topic. In this they have provided valuable insights to
their effectiveness and limitations[39, 4]. And also about the how AI-based systems effective in preventing cheating,
but also rise important concerns about privacy, accuracy and bias[31]. Security is the most important aspect of online
exam. To ensure the security of online examinations they proposed a continuous online authentication system[22].
This monitoring is done based on machine learning algorithms to find the cheating behavior of student[10]. In this
module if it finds any suspicious behavior it makes a record of it for viewing the real time administrators and it
terminates the exam[25]. Face recognition system by using eigen face method is used to monitor the students behavior
during online exams where face acts as priority index[11]. Eigen face consists of eigen vectors which are used in face
recognition[34]. With the help of webcam and microphone the teacher can monitor any student at any time. The system
can also keep the records of the malpractices[35]. The logs of malpractices are be used to verify the student manually
in case of suspicion. For developing this model they considered several features like face recognition, multiple person
detection, mobile detection and no person detection[33]. For face recognition they used the methodology called Local
Binary Pattern Histogram Algorithm. LBPH is a straight forward algorithm but very effective[18]. LBPH rate under
conditions of illumination, deflection of attitude and variability of expression is decreased[19]. It gives the output
as binary number. For mobile detection and persons detection they used the YOLO weights. It contains nearly 80
objects. It consists of 53 convolutional layers and each of them are followed by a batch normalization layer and
a RELU activation[2, 3]. Computer vision proposed as a solution for efficiently monitoring the student’s attention
through non invasive approach using Multi-task Convolutional Neural Network (MTCNN)[1]. The smart invigilation
system will be able to mark attendance of the student through face detection[15]. Cheating is the important concern
nowadays during online exams. To avoid cheating they developed a system by using webRTC that can stream audio
and video using twilio api for tracking the students behavior during the exam[38]. Server-less architecture by using
AWS was designed for monitoring the students during online examination. For this project face detection was done
by using EMTCNN model with convolutional network[24, 20]. Exams are the only ones that defines the students
learning capacity. From the time of covid many educational institutions are shifting to virtual mode, so that exams are
also conducting in online. In this phase cheating is the most concerned issue during online exams. To avoid this they
developed an AI agent that detects the cheating during exams[8].
The main idea of the project is to utilize the advanced technologies of computer vision methods and try to develop
an efficient and an effective online exam proctoring model for detecting the student’s attention during online exams[28,
7, 40].
The problem statement of the proposed work is to formulate an effective and user-friendly online exam proctoring
model that holds the potential to significantly revolutionize the educational evaluation process. The core aim is to
design an application capable of discerning vital cues such as body postures, object presence, audio cues, and potential
face spoofing. By systematically analyzing these indicators, the model seeks to make a decisive determination on
whether a student is engaging in any form of academic misconduct during the examination[36].
The objective of the work is to
 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408 1401
Tripty et al. / Procedia Computer Science 00 (2024) 000–000 3

• Detecting whether a student is engaging in copying during exams is typically a challenging task for teachers.
However, the introduction of our online exam proctoring model simplifies this process, enabling teachers to
readily identify instances of copying by students during exams.
• To find whether the student is copying or not is really hectic task and also it ma, if not proper systematic way is
followed. Our objective is to provide a platform for a teacher to make the student genuine during exam.
• To help teacher to identify whether student is copying or not. This application or recommendation system is
not only useful for the online exam proctoring, even it will be useful in class room monitoring. In some cases,
students try to cheat in exams, with the help of this application we can easily caught them during the exam.
• It can also provide real-time feedback to students and instructors. By monitoring students during the exam,
the system can identify areas where students are struggling and provide feedback and support to help them
improve. This objective is particularly important in online learning, where students may not have the same level
of interaction with their instructors as they would in traditional classroom settings.

2. Methodology

The proposed method utilizes a video input to identify whether the student in exam is cheating or not. Based on
that video data extract the features from video by using computer vision techniques. Utilizing identifiable features, as-
certain behavioral indicators, conduct testing, and subsequently predict whether a student taking an exam is engaging
in cheating behavior or adhering to academic honesty as shown in Fig. 1.

2.1. Proposed method

The proposed system architecture consists of an online exam proctoring model that can be able to identify the
students behavior during exam. It has 4 stages to develop an efficient online exam proctoring model. The stages are
Collecting data, Computed features, Behavioral clues and cheating status. For developing this model, utilized some
required python libraries and computer vision techniques such as cv2, mediapipe, tensorflow, pyaudio etc. Based on
the above mentioned techniques developed the efficient class room attention model as shown in Fig. 2. This phase
will encompass face recognition, face spoofing detection, and mouth tracking. It will address scenarios where multi-
ple persons are detected, focusing on these pivotal techniques crucial for effective online exam monitoring. Utilizing
OpenCV, the model will execute face detection, face spoofing detection, and mouth tracking. Tensorflow will be em-
ployed for the identification of multiple individuals, while Mediapipe will enable the detection of facial body postures.
Furthermore, the tensorflow framework will be leveraged to ascertain the count of individuals, and PyAudio will fa-
cilitate audio recognition. For object detection tasks, YOLO weights will be utilized to identify items such as mobile
devices[21]. Through the integration of these techniques, this model is being developed to achieve its intended ob-
jectives.Traditional in-person exams benefit from proctor supervision, which greatly reduces the chances of cheating.
However, in the context of distributed online exams, monitoring individual student behavior becomes a significant
challenge. Teachers struggle to assess each student’s exam situation and overall learning status. Additionally, the pro-
liferation of modern technology opens up various avenues for cheating, including text messaging, online searches, and
plagiarism, further complicating the integrity of online assessments[17].
The subsequent phase involves feature computation. In this step, features will be extracted based on the input
obtained from the initial stage. These features are integral for assessing students’ engagement during the exam. They
encompass extracted facial attributes, facial orientation, mouth tracking, audio recognition, the count of detected
individuals, and mobile device identification. The extraction of facial attributes and orientation will be carried out
through the utilization of the Mediapipe library, which facilitates the detection of a 32-point skeleton structure on
the face. For mouth tracking, the Scipy library will be employed to calculate the distance between the lips. This
distance will serve as a predictor of a person’s mouth status. Furthermore, this phase will encompass mobile device
detection and individual counting using the Tensorflow framework and YOLO weights. Audio recognition, on the
other hand, will be accomplished using the PyAudio library. The computed features within this phase encompass
facial attributes, facial orientation, mouth tracking, audio recognition, the count of detected individuals, and mobile
device detection. These computed features will serve as the foundation for establishing a comprehensive behavioral
classification system.
1402 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408
4 Tripty et al. / Procedia Computer Science 00 (2024) 000–000

Fig. 1: Block diagram of the proposed method

Following the feature computation stage, the subsequent task involves the creation of a behavioral classification
system. Within this behavioral class, the computed features will be categorized into two distinct classes: "Terminate
Exam" and "Continue Exam." For the "Continue Exam" class, the relevant features include the presence of a single
individual, active screen observation, head support, absence of mobile device detection, and direct gaze. Conversely,
the "Terminate Exam" class is characterized by multiple persons being detected, speech recognition triggers, instances
of looking away from the screen, mobile device identification, and indications of face spoofing. By evaluating these
specific features, the system will be adept at determining the status of academic integrity, discerning instances of
potential cheating during the exam.
 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408 1403
Tripty et al. / Procedia Computer Science 00 (2024) 000–000 5

Fig. 2: Block diagram of the proposed system architecture

2.1.1. YOLO
YOLOv3 (You Only Look Once, Version 3) stands as a real-time object detection algorithm designed to pinpoint
distinct objects within videos, live feeds, or images. Powered by the YOLO machine learning framework, this al-
gorithm harnesses features acquired through a deep convolutional neural network to accurately identify and detect
various objects[21, 12, 37, 9].
In the proposed work YOLO is utilized to identify the objects such as face of the student/examinee, Mouth of the
examinee, Audio detection, Mobile phone, Multiple person detection, head orientation of the examinee(left/right).

3. Result and Analysis

The proposed research encompassed several aspects including face orientation, face spoofing detection, mouth
tracking, audio recognition, mobile device detection, and person count estimation.

3.1. Front end part result

The work start with Face registration page window as shown in Fig. 3 and press s to capture the image of the
student. When attempting registration with either no individuals or multiple individuals, a message will be displayed
as "No face detected or Multiple faces detected. Please try again". Upon successful registration, the exam registration
page will become accessible as shown in Fig. 4. At this point, students are required to register using their respective
personal information. Once the exam registration is finished, the exam page will become accessible. Here, students are
expected to take the exam. Upon completion of the exam, the student will receive their respective score, as depicted
in Fig. 5.

3.2. Back end part result

It shows the features like face orientation, face spoofing, mouth tracking, audio detection, mobile detection, person
count as shown in proctoring window Fig. 15.
1404 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408
6 Tripty et al. / Procedia Computer Science 00 (2024) 000–000

Fig. 3: Registration page with image. Fig. 4: Exam registration page. Fig. 5: Exam Score.

Fig. 6: Face orientation. Fig. 7: Mouth Tracking.

3.2.1. Face orientation result

The system indicates the orientation of the student’s face, including positions like straight, left, right, up, or down
as shown in Fig. 6. In cases where the face orientation is not aligned properly, a message is displayed instructing the
student to focus on the screen.

3.2.2. Face Spoofing result

The system detects face spoofing in a student as shown in Fig. 15. This involves determining whether the individual
taking the exam is an actual person or an image. In the event that an image is detected, a message labeled "spoof" is
displayed.

3.2.3. Mouth tracking result

The system performs mouth tracking on the student’s face to entails monitoring whether the mouth is open or
closed as shown in Fig. 7.
 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408 1405
Tripty et al. / Procedia Computer Science 00 (2024) 000–000 7

Fig. 8: Audio Detection. Fig. 9: Mobile Detection.

3.2.4. Audio detection result

The system assesses whether the student is speaking by analyzing the incoming sound from the student’s side. If
any disruptions are detected in the audio, a message labeled "audio detected" is displayed as shown in Fig. 8.

3.2.5. Mobile detection result

The system is designed to detect the presence of mobile devices during the exam as shown in Fig 9. In case a mobile
device is detected while the exam is in progress, a message "mobile detected" is displayed. If the mobile device is
continuously detected for a duration of 10 seconds, the exam will be terminated.

3.2.6. Person count result

The system provides a count of the individuals present within the camera frame during the exam as displayed in
Fig.10. The person count is displayed on the proctoring window. If the system detects either multiple individuals or
no person within the frame continuously for a period of 10 seconds, the exam will be terminated.

3.3. System Testing

To determine whether our expected outcomes and actual results were the same, first established a few specific
cases and scenarios. To accomplish this, captured 70 movies and tested all of the defined cases and scenarios on those
videos.

3.3.1. The expected outcome match the actual result

As depicted in Fig. 11, it is evident that the video frame contains two individuals, with the second person observed
holding a mobile phone. The expected outcome was the detection of a mobile phone and the presence of multiple per-
sons, aligning with the actual outcome of detecting a mobile phone and identifying multiple persons. This congruence
between the expected and observed outcomes validates the successful passing of the test.
As depicted in Fig. 12, the video frame clearly illustrates an individual holding a mobile phone, with their head
positioned downwards as they engage with the device. As anticipated, the expected outcome encompassed the detec-
tion of an open mouth, the presence of a mobile phone, the head oriented to the right, and a request to focus on the
screen. Remarkably, the observed outcome mirrored these expectations precisely, thereby confirming the successful
passage of this particular test case. In Fig. 13, the visual representation indicates the individual holding a remote
control instead of a mobile phone. As projected, the expected outcome involved the identification of a remote control
rather than a mobile device. Optimistically, the actual outcome corresponded accurately by categorizing the object as
a remote control, thereby aligning with the anticipated result.
1406 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408
8 Tripty et al. / Procedia Computer Science 00 (2024) 000–000

Fig. 10: Person Count. Fig. 11: Mobile and multiple person testing.

Fig. 12: Head position, mouth status and mobile testing. Fig. 13: Object detection testing.

3.3.2. The expected outcome did not match the actual result
In Fig. 14, it can be observed that the individual within the video frame is not authentic; rather, an image was
positioned in front of the camera, constituting a spoof. As anticipated, the system correctly identified this as a spoof,
but unexpectedly, it also recognized the image as a genuine person, thereby erroneously increasing the person count
to 1.
In Fig. 15, it is evident that the individual depicted is not authentic; rather, it is an image. In such scenarios,
the expected response should label it as a spoof, resulting in a person count of 0. Furthermore, no head or mouth
status should be indicated. Surprisingly, the actual outcome deviated from this expectation. The system detected head
and mouth status, and contrary to the anticipated spoof classification, it identified the image as a genuine person,
consequently displaying a person count of 1.

4. Conclusion

Following the surge in demand caused by the ongoing COVID-19 Pandemic, online exams has emerged as the
subsequent trend for adoption, following in the footsteps of the already popular online learning. The proctor ensures
that there are no unethical practises occurring throughout the exam. The proposed work’s main goal is to make it
easier for teachers to determine whether a student has engaged in any sort of scamming during an exam. In order to
achieve this goal, created an online exam proctoring system that makes use of a facial recognition algorithm. This
algorithm serves a dual purpose: verifying the user’s identity and detecting any inappropriate behaviors exhibited by
 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408 1407

Tripty et al. / Procedia Computer Science 00 (2024) 000–000 9

Fig. 15: Spoof, person count and head orientation testing in the
Fig. 14: Spoof and person count testing. proctoring window.

the student throughout the online examination. The proposed model encompasses a comprehensive range of features,
including face recognition, identification of face spoofing attempts, monitoring of mouth movements, quantification
of detected individuals, identification of mobile device usage, and even audio analysis.
Advanced AI algorithms and larger feature set will produce a more reliable and accurate system, improving its
proctoring and exam security abilities.

References

[1] Akalanka, P., Manathunga, K., 2022. Real-time exam anomaly detection in moodle-based exam systems with an ai agent, in: 2022 International
Research Conference on Smart Computing and Systems Engineering (SCSE), IEEE. pp. 217–224.
[2] Alioua, N., Amine, A., Rogozan, A., Bensrhair, A., Rziza, M., 2016. Driver head pose estimation using efficient descriptor fusion. EURASIP
Journal on Image and Video Processing 2016, 1–14.
[3] Arnò, S., Galassi, A., Tommasi, M., Saggino, A., Vittorini, P., 2021. State-of-the-art of commercial proctoring systems and their use in
academic online exams. International Journal of Distance Education Technologies (IJDET) 19, 55–76.
[4] Asep, H.S., 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), IEEE. pp. 284–289.
[5] Babu, T., Gupta, D., Singh, T., Hameed, S., Zakariah, M., Alotaibi, Y., 2021a. Robust magnification independent colon biopsy grading system
over multiple data sources. Computers, Materials Continua 69, 99–128. doi:10.32604/cmc.2021.016341.
[6] Babu, T., Nair, R.R., 2023. Colon cancer prediction with transfer learning and k-means clustering, in: Frontiers of ICT in Healthcare: Proceed-
ings of EAIT 2022. Springer, pp. 191–200.
[7] Babu, T., Singh, T., Gupta, D., Hameed, S., 2021b. Colon cancer prediction on histological images using deep learning features and bayesian
optimized svm. Journal of Intelligent Fuzzy Systems 41, 1–12. doi:10.3233/JIFS-189850.
[8] Canedo, D., Trifan, A., Neves, A.J., 2018. Monitoring students’ attention in a classroom through computer vision, in: Highlights of Practical
Applications of Agents, Multi-Agent Systems, and Complexity: The PAAMS Collection: International Workshops of PAAMS 2018, Toledo,
Spain, June 20–22, 2018, Proceedings 16, Springer. pp. 371–378.
[9] Chen, Y., Yuan, X., Wu, R., Wang, J., Hou, Q., Cheng, M.M., 2023. Yolo-ms: Rethinking multi-scale representation learning for real-time
object detection. arXiv preprint arXiv:2308.05480 .
[10] Coghlan, S., Miller, T., Paterson, J., 2020. Good proctor or" big brother"? ai ethics and online exam supervision technologies. arXiv preprint
arXiv:2011.07647 .
[11] Dadashzadeh, M., 2021. The online examination dilemma: To proctor or not to proctor?. Journal of Instructional Pedagogies 25.
[12] Diwan, T., Anirudh, G., Tembhurne, J.V., 2023. Object detection using yolo: Challenges, architectural successors, datasets and applications.
multimedia Tools and Applications 82, 9243–9275.
[13] Dixit, A.K., Nair, R.R., Babu, T., 2022. Analysis and classification of restaurants based on rating with xgboost model, in: 2022 3rd International
Conference on Issues and Challenges in Intelligent Computing Techniques (ICICT), IEEE. pp. 1–6.
[14] Gayathri, R., Pati, P.B., Singh, T., Nair, R.R., 2022. A framework for the prediction of diabtetes mellitus using hyper-parameter tuned xgboost
classifier, in: 2022 13th International Conference on Computing Communication and Networking Technologies (ICCCNT), IEEE. pp. 1–5.
[15] Ghizlane, M., Hicham, B., Reda, F.H., 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), IEEE. pp. 1–5.
1408 Tripty Singh et al. / Procedia Computer Science 235 (2024) 1399–1408
10 Tripty et al. / Procedia Computer Science 00 (2024) 000–000

[16] Haarika, R., Babu, T., Nair, R.R., 2023. Insect classification framework based on a novel fusion of high-level and shallow features. Procedia
Computer Science 218, 338–347.
[17] Jia, J., He, Y., 2022. The design, implementation and pilot application of an intelligent online proctoring system for online exams. Interactive
Technology and Smart Education 19, 112–120.
[18] Kaddoura, S., Popescu, D.E., Hemanth, J.D., 2022. A systematic review on machine learning models for online learning and examination
systems. PeerJ Computer Science 8, e986.
[19] Kulshrestha, A., Gupta, A., Singh, U., Sharma, A., Shukla, A., Gautam, R., Kumar, P., Pandey, D., 2023. Ai-based exam proctoring system,
in: 2023 International Conference on Disruptive Technologies (ICDT), IEEE. pp. 594–597.
[20] Lee, K., Fanguy, M., 2022. Online exam proctoring technologies: Educational innovation or deterioration? British Journal of Educational
Technology 53, 475–490.
[21] Li, C., Wang, R., Li, J., Fei, L., 2020. Face detection based on yolov3, in: Recent Trends in Intelligent Computing, Communication and
Devices: Proceedings of ICCD 2018, Springer. pp. 277–284.
[22] Mitra, A., 2023. Redesign of online proctored exams for stem learners in heis: proposal for incorporating higher-order thinking skills and
democratic pedagogy via operhot platform. FEMS Microbiology Letters , fnad074.
[23] Mohan, V., Gowda, A., Nair, R.R., Kishore, S., Babu, T., 2023. Face mask detection using mask r-cnn to control the spread of covid-19, in:
2023 International Conference on Recent Trends in Electronics and Communication (ICRTEC), IEEE. pp. 1–6.
[24] Monteiro, S., Bhate, R., Sharma, L., Shaikh, P., 2022. Proct-xam–ai based proctoring, in: 2022 2nd Asian Conference on Innovation in
Technology (ASIANCON), IEEE. pp. 1–6.
[25] Motwani, S., Nagpal, C., Motwani, M., Nagdev, N., Yeole, A., 2021. Ai-based proctoring system for online tests, in: Proceedings of the 4th
International Conference on Advances in Science & Technology (ICAST2021).
[26] Nair, R.R., 2019. Multi-sensor medical image fusion using pyramid-based dwt: a multi-resolution approach. IET Image Processing 13,
1447–1459(12). URL: https://digital-library.theiet.org/content/journals/10.1049/iet-ipr.2018.6556.
[27] Nair, R.R., Babu, T., Kishore, S., 2023a. Recent and emerging technologies in industrial iot. Opportunities and Challenges of Industrial IoT in
5G and 6G Networks , 50–72.
[28] Nair, R.R., Babu, T., Singh, T., 2023b. Multiresolution approach on medical image fusion by modified local energy. Signal, Image and Video
Processing , 1–8.
[29] Nair, R.R., Singh, T., 2021a. Mamif: multimodal adaptive medical image fusion based on b-spline registration and non-subsampled shearlet
transform. Multimedia Tools and Applications 80, 19079–19105.
[30] Nair, R.R., Singh, T., 2021b. An optimal registration on shearlet domain with novel weighted energy fusion for multi-modal medical images.
Optik 225, 165742. URL: https://www.sciencedirect.com/science/article/pii/S0030402620315692, doi:https://doi.org/
10.1016/j.ijleo.2020.165742.
[31] Nigam, A., Pasricha, R., Singh, T., Churi, P., 2021. A systematic review on ai-based proctoring systems: Past, present and future. Education
and Information Technologies 26, 6421–6445.
[32] Nurpeisova, A., Shaushenova, A., Mutalova, Z., Ongarbayeva, M., Niyazbekova, S., Bekenova, A., Zhumaliyeva, L., Zhumasseitova, S., 2023.
Research on the development of a proctoring system for conducting online exams in kazakhstan. Computation 11, 120.
[33] Selwyn, N., O’Neill, C., Smith, G., Andrejevic, M., Gu, X., 2023. A necessary evil? the rise of online exam proctoring in australian universities.
Media International Australia 186, 149–164.
[34] Sorate, P., Vpkbiet, B., Chhajed, G., 2017. Survey paper on eye gaze tracking methods and techniques. International Research Journal of
Engineering and Technology 4, 465–468.
[35] Sridhar, A., Rajshekhar, J., 2022. Ai-integrated proctoring system for online exams. Journal of Artificial Intelligence 4, 139–148.
[36] Stocco, J.A., Otsuka, J.L., Beder, D.M., Micheloti, G.C., de Araujo, R.B., 2012. Logical and physical sensors-based online assessment security
support, in: 2012 IEEE 12th International Conference on Advanced Learning Technologies, IEEE. pp. 410–412.
[37] Terven, J., Cordova-Esparza, D., 2023. A comprehensive review of yolo: From yolov1 to yolov8 and beyond. arXiv preprint arXiv:2304.00501
.
[38] Thampan, N., Arumugam, S., et al., 2022. Smart online exam invigilation using ai based facial detection and recognition algorithms, in: 2022
2nd Asian Conference on Innovation in Technology (ASIANCON), IEEE. pp. 1–6.
[39] Thombare, C., Sapate, K., Rane, A., Hutke, A., 2022. Exam proctoring system. Int. J. Res. Appl. Sci. Eng. Technol 10, 466–470.
[40] Tripathi, A., Basavapattana, A., Nair, R.R., Singh, T., 2021. Visualization of covid bimodal scan using dnn, in: 2021 12th International
Conference on Computing Communication and Networking Technologies (ICCCNT), IEEE. pp. 01–07.