2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE)

EDUGUARD – AI BASED ONLINE EXAM

2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE) | 979-8-3315-9709-2/25/$31.00 ©2025 IEEE | DOI: 10.1109/ICCRTEE64519.2025.11052922

MANAGEMENT SYSTEM
Dr. R. Raja Subramanian Brundavanam Surekha Peram SubbaRao
Department of Computer Science and Department of Computer Science and Department of Computer Science and
Engineering Engineering Engineering
Kalasalingam Academy of Research and Kalasalingam Academy of Research and Kalasalingam Academy of Research
Education Education and Education
Virudhunagar, Tamilnadu, India Virudhunagar, Tamilnadu, India Virudhunagar, Tamilnadu, India
[email protected] [email protected] [email protected]

G. Adi Kesava Chowdary M. Venkatesh

Department of Computer Science and Department of Computer Science and
Engineering Engineering
Kalasalingam Academy of Research and Kalasalingam Academy of Research and
Education Education
Virudhunagar, Tamilnadu, India Virudhunagar, Tamilnadu, India
[email protected] [email protected]

The growth of distance learning and digital education has brought to

Abstract— EduGuard Secured Exam Portal is a complete, secure light a need for effective, secure, and scalable examination systems.
and automated online examination system that resolves all issues Traditional examination methods which are primarily dependent on
of registered online and traditional assessments including cheating, manual invigilation, paper-based assessments, and human grading, are
grading mistakes, etc. With such a system, all these tasks become tedious, time-consuming, prone to errors, and offer opportunities for
complete in the form of an AI powered proctoring, automated academic dishonesty. With the increase of online education, educational
grading, real time monitoring and RBAC (Role Based Access institutions are increasingly struggling with the maintaining of exam
Controls) to make exams secure, intangible as well as efficient. It integrity, administrative burden, and fair assessment.
supports multiple sorts of question formats: MCQs, short answer
and long answers questions are all supported for flexible and multi The EduGuard Secured Exam Portal or Online Exam Management
phased tests. Next is AI Proctoring, based on ML5.js which is an System (OEMS) tackles these challenges by automating processes and
ongoing exam monitor of students in detecting student behaviour integrating artificial intelligence (AI) technologies, and contemporary
that is suspicious like it is using the unauthorized device, multiple security protocols. Apart from these issues, schools and universities are
faces, background noise etc which results into less cheating. All that challenged by exam logistics, manual grading, and the inability to scale
is developed in this system using PHP, MySQL, JavaScript and examinations. The shift to online learning environments has
exacerbated these problems and called for a strong, non-compromising
Tailwind CSS. It is secured with AES-256 encryption of sensitive
technological solution that preserves the integrity of the assessment
exam data so faculty work is cut in half doing automatic grading as
while decreasing the administrative burden.
a way to provide instantaneous feedback to students. Such a
performance analysis features allow educators to inspect great
detail of results of exams and identify the student performance Current online examination systems do not provide real time proctoring
trends. The use of such a system allows a 90 % success rate in with AI. These systems are robbery prone to cheating. If the answers
are subjective, then there is no automated grading, and thus there are
cheating detection and a 70 % reduction in grading time, and
inconsistencies. Lacking strong multi layered security feature, this
therefore, this becomes feasible solution for any educational
poses a higher chance of the fraudsters or hackers using such unsecured
institution. Since it is modular architecture, it is easy to integrate means for their gains. Assistance to academic integrity and fairness can
with Learning Management Systems (LMS) and have some other be provided by the enhancement of AI monitoring, automated
advanced proctoring & blockchain based security or anti tamper evaluation, and robust security.
proof record of the exam and some other accessibility features.
EduGuard is a ground breaking product that allows for the
seamless automation of any education establishment exhausting • The system uses ML5.js to monitor students in real time,
the need of inspecting and securing them using advanced teaching detecting suspicious activities like multiple faces, external
and stepping up the efficiency of tests as well as keeping the devices, or unusual behaviour to prevent cheating.
academic honesty in virtual education. • The system automatically scores multiple-choice and short-
answer questions, while keyword-based semantic analysis
Index Terms— AI Proctoring, Automated Grading, Exam helps assess long-answer responses, reducing grading time by
70%.
Security, Cheating Detection, Online Examination System.
• Features like AES-256 encryption and Role-Based Access
Control (RBAC) ensure secure authentication, preventing
unauthorized access and protecting exam data.

I. INTRODUCTION

979-8-3315-9709-2/25/$31.00 ©2025 IEEE

Authorized licensed use limited to: KJ's Educational Institute. Downloaded on July 24,2025 at 06:08:51 UTC from IEEE Xplore. Restrictions apply.
 2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE)
• A discriminator-based AI system has a 90% success rate in Based on these AI driven real time proctoring, automated objective
detecting dishonest behaviour and considerably enhances the and subjective grading, the foundation of the Proposed Online Exam
integrity of exams. Management System (OEMS) is construction on AES-256
• We plan to integrate with Learning Management Systems encryption, Role Based Access Control (RBAC). This paper focuses
(LMS) for an academic use and utilize blockchain technology on which gaps in online assessment technology OEMs intends to
for more security and reliability purposes. close: scalability, exam integrity and usability.

II. BACKGROUND & RELATED WORKS REFEREN TECHNIQ KEY LIMITATIO

CE UE USED CONTRIBUTI NS
First generation online examination systems had the focus on very ON
basic functionality like the structure of role-based access and web-
based delivery. Ji Wen [1] PHP & PHP & PHP &
MySQL- MySQL-based MySQL-
Over the years, online examination systems have made remarkable based system based system
progress which has tackled many problems involving scalability, system
automation, security and proctoring. Then, in 2001, Ji Wen [1] Huiqiang Lu J2EE-based J2EE-based J2EE-based
created an early PHP and MySQL based exam system that had [2] online exam online exam online exam
structured roles for students, teachers, and administrators. The system platform platform platform
was, however, not very flexible and scalable. Huiqiang Lu [2] Chen
proposed an online exam platform based on J2EE providing better Web-based Web-based Web-based
Xiangjun [3] general general exam general exam
performance for the online environment, serving a large number of
exam system system
users.
system
Sophal Chao Teacher- Teacher- Teacher-
Chen Xiangjun [3] also further improved general purpose exam
[4] controlled controlled controlled
management by improving how question banks were organized as
well as the accessibility for those using the system. Instead, Sophal exams exams exams
Chao [4] examined teacher-controlled examining (i.e. exams) and JunJie Zhang Template- Template-based Template-
highlighted the advantages of this for institutions to gain control over [5] based intelligent based
entire departments and information sharing of exam administration. intelligent assessments intelligent
Though these progess had been made, automated proctoring and assessments assessments
security lacked, continuing to cause concerns with respect to exam G. Vikram ASP.NET- ASP.NET-based ASP.NET-
integrity. [6] based failure recovery based failure
failure recovery
The intelligent online testing is developed by JunJie Zhang [5] that is recovery
based on the template driven assessment to enhance the assessment Lio Siyao [7] Cheating Cheating Cheating
efficiency for adaptive learning. Student responses to exams were the technique technique technique
basis of this approach which personalized exams, but we still needed analysis analysis analysis
better anti-cheating mechanisms. G. Vikram [6] addressed another
important issue, that is, to reinstate students in ASP.NET based exams Monica Automated Automated Automated
on disconnections. Nevertheless, these systems did not have Farrow [8] programmi programming programming
sophisticated proctoring techniques to prevent unfair practices. ng evaluation evaluation
evaluation
Lio Siyao [7] evaluated the common online cheating techniques and Guo Ping Automated Improved No support for
found in the security loophole what they used system vulnerabilities [9] scoring for grading long-answer
through screen sharing, external devices and impersonation. The MCQs and efficiency and grading
outcome of this research was that of the need of AI driven monitoring short answers accuracy
and behavioural tracking. Another programming research paper
machine developed by Monica Farrow [8] automatically justified the Afzaal Power failure Allowed students Did not
redundancies of the Java based programming evaluation in the Ahmad et al. handling & to revisit flagged integrate AI-
technical class. In a similar attempt of applying an online grading [10] Dropbox- questions before based security
system based on scoring system for Multiple choice and short answer style submission or proctoring
questions, it is also helpful to decrease the burden on educators as question
stated by Guo Ping [9]. flagging

The first introduction of power failure handling and a Dropbox like

question flagging for PHP and MySQL based online exams was made III. METHODOLOGY
by Afzaal Ahmad and his colleagues [10] in response to the need for
The study applies a mixed-methods where qualitative and
more flexibility. This made it better for users, though there were no
quantitative research are conducted to evaluate the Online Exam
AI powered proctoring, subjective grading automation or full security
Management System (OEMS) on effectiveness. User response in the
in tact on existing platforms.
form of surveys is qualitative, whereas quantitative is system

Authorized licensed use limited to: KJ's Educational Institute. Downloaded on July 24,2025 at 06:08:51 UTC from IEEE Xplore. Restrictions apply.
 2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE)
performance measures like marking accuracy, rate of detection of introduced into the online examination by such automated grading
cheating, and response time. systems.

A. Data Collection Methods:

Surveys: Feedback from teachers, students, and administrators is Equation (3) calculates the System Accuracy (SA), that is algebraic
collected in an effort to learn about usability, accessibility, and measure of the advisability of the AI based grading system. The
satisfaction with the system. Surveys consist of closed-ended (e.g., number of correctly graded answers AcA_cAc and the number of
Likert scale) and open-ended questions. answers evaluated TaT_aTa are represented by AcA_cAc and
TaT_aTa, respectively. And this equation is to assess the reliability
Experiments: The system is pilot-tested in a real education of automated grading for the fair evaluation of the students'
environment to observe how it performs. Grading time taken, how response.
good the cheating is detected, and if and when the system is present
are quantified.

Datasets: Training and cross-validation utilize available datasets of

exam questions, student answers, and instances of cheating. Equation (4) Shannon entropy is used to calculate the Encryption
Strength (H), to measure the level of security on AES-256
B. Tools, Algorithms, or Models Used encryption. The key length (LLL) is 256 bits (256 bits used for AES-
256) and NNN is the number of possible values per bit (2 for binary
Backend Development: PHP is utilized in server-side development encryption). This equation represents the robustness of encryption
to simplify frontend to database communication. and its ability to keep exam data secure and prevents unauthorized
access to them.
Database Management: MySQL is employed for managing exam
data, users, and scores. Database is being normalized for effective
data management and data integrity.

Frontend UI Design: Tailwind CSS for user interface and responsive

Equation (5) This calculates the User Load Scalability (UL), which
design.
is the share of capacity being used by the system. Since the number
of active users a system supports is limited, uaU_aUa is the number
AI Proctoring: ML5.js, an ML library, for real-time object detection
of active users a system has at a certain time, and umU_mUm is the
in order to monitor the students under surveillance while
maximum number of users a system can support. Evaluation of the
administering exams. The model detects suspicious behaviours such
system’s capability to cope with concurrent users without
as other devices or multiple faces.
performance degradation in using this equation.
Automated Marking: Rule-based systems are employed for
marking multiple-choice and short-answer questions. Keyword
matching and semantic analysis are applied for long-answer
questions.
Equation (6) It calculates Exam Session Integrity Score (ESI), a
Security Controls: Some features are provided access to only score of the reliability of AI proctoring in ensuring exam integrity.
privileged users (admins, teachers, students) via Role-Based Access number of detected violations is represented by VdV_dVd, and total
Control (RBAC). Data encryption (AES-256) is provided to encrypt number of monitored sessions as VtV_tVt. Answers the question by
sensitive information. this equation of helping to analyze the effectiveness of security
measures in online exams to reduce dishonest activities.
C. Equations
D. Model procedure and Evaluation

Creation of Exam: Teachers provide the information on exam

name, length and question type (multiple choice, short answer or
Equation (1) The Cheating Detection Rate (CDR) is calculated as the
long answer) for teachers to develop the exam.
conventional proportional of the cheating detected by the AI
proctoring system. Thus, here, DcD_cDc represent the number of
Real-Time Proctoring: This monitoring software proctors the user
detected cheating instances and TcT_cTc be the total number of
by being able to see their webcam and test screen. Alerts are raised
cheating attempts. In order to assess the extent of the online exam
on behaviour, and the system keeps evidence and saves it for
proctoring using AI, this equation is used.
analyzing.

Automated Grading: This involves the system automatically

grading multiple choice as well as short answer questions. With
these long answer questions, the instructor can grade them and give
Equation (2) The Grading Time Reduction (GTR) is calculated – certain feedback.
the percentage decrease in grading time as a result of automation.
The manual grading time is Tm-Ta/Tm. But this equation is Result Management: Results are stored in the database to be retrieved
essential for revealing how much the improvements of efficiency

Authorized licensed use limited to: KJ's Educational Institute. Downloaded on July 24,2025 at 06:08:51 UTC from IEEE Xplore. Restrictions apply.
 2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE)
by teachers as well students. Question-wise response, i.e., extensive
C. Database Operations: The database stores and retrieves data as
analysis and average score. requested by the backend. Exam results are stored and can be
accessed by authorized users.
Security: The defence of any unauthorized access is done by proper
authentication (self. and/or session management). D. AI-Based Proctoring: During exams, the AI-based proctoring
system monitors students in real-time. Suspicious activities are
E. Ethical Considerations: flagged and recorded for review.

Data Privacy: The user information, such as the exam answers and E. Security: All data exchanges between the frontend, backend, and
personal data, are stored and encrypted securely. The sensitive data database are encrypted. User roles and permissions are enforced to
are viewed by the concerned users only. ensure secure access

Informed Consent: The clients are made aware of the surveillance System Flow Diagram: The system flow diagram represents the
feature of the system, and consent is obtained before enabling the step-by-step flow of processes within the system.
webcam for proctoring.
F. Figures and Tables
Transparency: The system clearly indicates what constitutes
acceptable conduct while taking an exam and what amounts to
cheating and consequent penalty.

Elimination of bias: The model that is based on AI is trained on full

sets of data so as not to be biased and to ensure fair monitoring to all.

Conformity: The system adheres to the institution's ethics codes as

well as data protection legislations such as GDPR and FERPA.

Libraries and Frameworks: Online Exam Management System

(OEMS) has leveraged the use of libraries and frameworks to achieve
a responsive user interface, scalability, and performance.

VI. SYSTEM ARCHITECTURE

Fig 2: Admin to create user accounts

Fig 1: System Flow Diagram Fig 3: Teacher create exam & manage exams

A. User Interaction: Students, teachers, and administrators interact

with the system through the frontend. Requests (e.g., login, exam
creation, result viewing) are sent to the backend.

B. Backend Processing: The backend processes requests, performs

necessary validations, and interacts with the database. For exams, the
backend retrieves questions, stores answers, and triggers the AI-based
proctoring system.

Authorized licensed use limited to: KJ's Educational Institute. Downloaded on July 24,2025 at 06:08:51 UTC from IEEE Xplore. Restrictions apply.
 2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE)

Fig 4: Teacher to Add question and manage question

Fig 9. Tables in MYSQL Database (In phpMyAdmin)

Fig 5: Students write the exam under rules and restrictions

Fig 10: UML diagram of database design

V. RESULT ANALYSATION

The Online Exam Management System (OEMS) demonstrates

significant improvements in examination integrity, efficiency, and
security. The AI-powered proctoring system achieved a 90%
success rate in detecting cheating, significantly reducing academic
malpractice. The automated grading system streamlined the
evaluation process, cutting grading time by 70%, thereby reducing
Fig 6: Verify the user results and publish results the administrative burden on educators.

From a security perspective, the integration of AES-256 encryption

and Role-Based Access Control (RBAC) ensured secure user
authentication and data protection, minimizing unauthorized access
risks. The system's scalability allowed it to handle a large number
of concurrent users, making it suitable for institutions of various
sizes.
Fig 7: List of exams in student portal
The users rated the usage and efficiency highly. The students found
real time proctoring to be a blessing and it was of great help in
keeping the exam integrity and for the teachers accessing automated
grading feature was also a thing which was found beneficial.
However, users suggested that even with all of that, humans
screwed it up, and proposed that more could be done to help get
better separation between positives and negatives in detection of
cheating from AI proctoring.
Fig 8: Student exam results
Future enhancements to the system are suggested for further
improvement in the system’s performance, namely, future integration on
the Learning Management System (LMS) and with other advanced AI

Authorized licensed use limited to: KJ's Educational Institute. Downloaded on July 24,2025 at 06:08:51 UTC from IEEE Xplore. Restrictions apply.
 2025 International Conference on Computational Robotics, Testing and Engineering Evaluation (ICCRTEE)
models to detect cheating. We are also evaluating whether it could be [8] M. Farrow and P. J. B. King, "Experiences with online
used to create a more secure and less likely susceptible system to fraud. programming examinations," IEEE Trans. Educ., vol. 51, no.
2, pp. 251-255, May 2008.
Overall, the OEMS has significantly delivered on the promise of targeted [9] G. Ping, Z. Hai-yan, and D. Yiwen, "The design and
innovations addressing a number of online exams challenges and implementation of online scoring system," in Proc. IEEE Int.
provides an effective signal for additional enhancements. Symp. IT Med. Educ. (ITIME), vol. 1, 2009, pp. 606-610.
[10] A. Ahmad, N. U. Khan, and A. W. Abbas, "PHP+MySQL
based online examination system with power failure handling
VI. CONCLUSION
and Dropbox capability," in Proc. 7th IEEE Int. Conf. Softw.
Security Rel. Companion (SERE-C), Gaithersburg, MD,
The Online Exam Management System (OEMS) has succeeded in USA, Jun. 2013, pp. 27. doi: 10.1109/SERE-C.2013.27
entering the automated grading system with artificial intelligence based
protecting and sophisticated security features, due to the inability of
traditional methods of examination to address the speculated
deficiencies. The training and admin working for this system uses PHP,
MySQL and ML5.js for development in order to reduce admin workload
and increase security and fairness in exams. The experimental results
indicate that it succeeds in 90% of cheating detection cases and that
result is very effective in reducing the grading time by 70%, which
makes it an ideal solution of modern educational institutions.

It has a scalable architecture to support the users of small institutions as

well as large scale Universities, with no degradation of performance.
Security features such as AES 256 encryption and role-based access
control are ways of ensuring confidentiality and compliance of a data
privacy regulation. Finally, ethical imperative, including transparency,
fairness, and informed consent was brought onto the board to create trust
in the examination where no existed.

The integration with LMS will emerge in the future work that will enable
the creation of a full educational ecosystem. Further enhancements will
be done to AI driven proctoring via incorporating more advanced
behavioural analysis techniques to increase the cheating detection
accuracy. Additionally, we will delve deeper into analysing blockchain
based security with which the data integrity is safeguarded and
tampering is prevented and lastly giving the audit trail for trace of
examination records.

OEMS does this with automation, security and scalability, and

establishes a new standard for digital assessment systems.

VII. REFERENCES

[1] J. Wen, Y. Zang, S. Xia, and Y. Yan, "A web-based examination system
based on PHP+Mysql," in Proc. 27th Annu. Int. Conf. Eng. Med. Biol.
Soc. (IEEE-EMBS), 2005, pp. 2882-2885.
[2] H. Lu and Y. Hu, "The design and implementation of online examination
system based on J2EE," in Proc. Int. Conf. Ind. Control Electron. Eng.
(ICICEE), Aug. 2012, pp. 93-95.
[3] C. Xiangjun and W. Fangshang, "Web-based general examination
system development," in Proc. 2nd Int. Conf. Comput. Eng. Technol.
(ICCET), vol. 3, Apr. 2010, pp. V3-457-V3-461.
[4] S. Chao and Y. B. Reddy, "Online Examination," in Proc. 5th Int. Conf.
Inf. Technol. New Gener. (ITNG), Apr. 2008, pp. 1266-1267.
[5] J. Zhang, W. Fang, and J. Song, "The design and realization of the
intelligentize online testing system based on templates," in Proc. 2nd Int.
Symp. Knowl. Acquis. Model. (KAM), vol. 3, Nov. 2009, pp. 248-251.
[6] G. Vikram and M. V. Vijaya, "Handling network failure in web-based
online exam," J. Current Comput. Sci. Technol., vol. 1, no. 7, pp. 407-
413, 2011.
[7] L. Sioyao and G. Qianrang, "The research on anti-cheating
strategy of online examination system," in Proc. 2nd Int. Conf.
Artif. Intell. Manage. Sci. Electron. Commerce (AIMSEC),
Aug. 2011, pp. 1738-1741.

Authorized licensed use limited to: KJ's Educational Institute. Downloaded on July 24,2025 at 06:08:51 UTC from IEEE Xplore. Restrictions apply.