AUTOMATED EXAMINATION INTEGRITY

MONITORING SYSTEM

A PROJECT REPORT

Submitted by

SARGURURAMAN - 212621104055
TAMILSELVAN M - 212621104067
TEJ SABAREESH V - 212621104068

in partial fulfillment for the award of the degree of

BACHELOR OF ENGINEERING

IN

COMPUTER SCIENCE AND ENGINEERING

SRI MUTHUKUMARAN INSTITUTE OF TECHNOLOGY,

CHENNAI-69

ANNA UNIVERSITY: CHENNAI 600 025

MAY 2025
AUTOMATED EXAMINATION INTEGRITY
MONITORING SYSTEM

A PROJECT REPORT

Submitted by

SARGURURAMAN - 212621104055
TAMILSELVAN M - 212621104067
TEJ SABAREESH V - 212621104068

in partial fulfillment for the award of the degree of

BACHELOR OF ENGINEERING

IN

COMPUTER SCIENCE AND ENGINEERING

SRI MUTHUKUMARAN INSTITUTE OF TECHNOLOGY,

CHENNAI-69

ANNA UNIVERSITY: CHENNAI 600 025

MAY 2025

i
 BONAFIDE CERTIFICATE

Certified that this Report titled “AUTOMATED EXAMINATION

INTEGRITY MONITORING SYSTEM is the bonafide work of
SARGURURAMAN N,TEJ SABAREESH V,TAMILSELVAN M
(212621104055 , 212621104068 , 212621104067) who carried out the work
under my supervision. Certified further that to the best of my knowledge the
work reported herein does not form part of any other thesis or dissertation on
the basis of which a degree or award was conferred on an earlier occasion on
this or any other candidate.

SIGNATURE SIGNATURE
Dr.D.RAJINIGIRINATH,M.Tech.,Ph.D. Dr.B.ASRAF YASMIN,MCA,M.Phil,Ph.D
HEAD OF THE DEPARTMENT SUPERVISOR
Professor Assistant Professor
Department of Computer Science Department of Computer Science and
and Engineering Engineering
Sri Muthukumaran Institute of Sri Muthukumaran Institute of
Technology, Technology,
Chikkarayapuram, Chennai-600 069 Chikkarayapuram, Chennai-600 069

Submitted for the Anna University Project Viva Voice held on ______________

INTERNAL EXAMINER EXTERNAL EXAMINER

ii
 ACKNOWLEDGEMENT

We are immensely grateful to our Chairperson & Managing Trustee,

Mrs.Gomathi Radhakrishnan, for her encouragement and guidance. We would
also like to extend our sincere gratitude to Dr.K.P.Gautham Srinivas, Chairman
and Dr.K.P.Arvind Srinivas, Managing Director, for their unwavering support and
motivation, which played a significant role in the successful completion of our
project.

We wish to express our sincere gratitude to our Principal

Dr.K.Somasundaram, for providing us an adequate infrastructure and congenial
academic environment and We thank our respected Vice Principal Dr.V.Anitha for
her encouragement and support in doing the project.

We express our deepest gratitude to our respectable Head of the Department,

Project Coordinator Dr.D.Rajinigirinath, M.Tech.,Ph.D., Professor of Computer
Science and Engineering and Project Supervisor Dr.B.Asraf Yasmin, Assistant
Professor,for their special interest on this project and giving this consistent support
and guidance during all the stages of this project work.

Finally, we thank all the teaching and non-teaching staff members of the
Department of Computer Science and Engineering of our college who helped us to
complete this project.

Above all we thank our parents and our family members for their constant
support and encouragement for completing this project.

iii
 ABSTRACT
With the growing prevalence of online education and remote assessments,
ensuring the integrity of examinations has become a significant challenge for
educational institutions, certification bodies, and training organizations. Traditional
proctoring methods are often resource-intensive, prone to human error, and difficult
to scale across large candidate pools. To address these limitations, this paper
proposes an Automated Examination Integrity Monitoring System (AEIMS) a
comprehensive solution that combines artificial intelligence (AI), computer vision,
and behavioral analytics to monitor and safeguard the credibility of examinations in
real time.The AEIMS architecture integrates multiple monitoring layers including
webcam surveillance, screen activity tracking, biometric verification, and audio
analysis. Facial recognition and liveness detection ensure the authenticated
candidate remains present throughout the examination. Eye-tracking algorithms
monitor focus and detect off-screen glances, while microphone input is analyzed to
identify unauthorized speech or background noise indicative of collusion. Screen
recording and keystroke logging further help in detecting the use of external
software, internet searches, or file access during the test session.To enhance
adaptability, the system employs machine learning techniques that continuously
train on new behavioral data, improving the accuracy of anomaly detection and
reducing false positives. The solution also includes a centralized dashboard for
administrators, providing real-time alerts, risk scores, and post-exam audit logs for
detailed review and action.Extensive testing in controlled environments and real-
world examination scenarios shows that AEIMS significantly improves the
reliability and security of digital assessments. It reduces reliance on human
invigilation, lowers operational costs, and supports large-scale deployment across
multiple locations. Furthermore, the system is designed to comply with privacy
regulations, incorporating data encryption, anonymization, and secure storage
practices.
iv
 TABLE OF CONTENT

CHAPTE PAGE
TITLE
R NO
ABSTRACT iv
LIST OF FIGURES viii
LIST OF TABLES ix
LIST OF ABBRIVATIONS x
1. INTRODUCTION 2
1.1 BACKGROUND AND MOTIVATION 2
1.2 PROBLEM STATEMENT 3
1.3 OBJECTIVES OF THE STUDY 4
1.4 SIGNIFICANCE AND SCOPE OF THE PROJECT 4
2. EXISTING SYSTEM 7
2.1 OVERVIEW OF CURRENTLY EXISTING SYSTEM 7
2.2 LIMITATIONS OF EXISTING PLATFORMS 8
2.3 USER FEEDBACK AND MARKET GAPS 9
2.4 LITERATURE REVIEW 10
2.4.1 TECHNOLOGICAL APPROACHES AND
10
EFFICENCY
2.4.2 ETHICAL AND PRIVACY CONCERNS 11
2.4.3 STUDENT EXPERIENCE AND
11
PSYCHOLOGICAL IMPACT
3 PROPOSED SYSTEM 14
3.1 SYSTEM OVERVIEW 14
3.1.1 PROJECT SUMMARY AND CONCEPT 15
3.1.2 KEY FEATURES AND FUNCTIONALITIES 16

v
 3.1.3 BENEFITS OVER EXISTING SYSTEMS 18
3.1.4 USE CASE SCENARIOS AND APPLICATIONS 19
3.2 SYSTEM ARCHITECTURE 20
3.2.1 LAYERED ARCHITECTURE DESCRIPTION 21
3.2.2 COMPONENT INTERACTION DIAGRAM 21
3.2.3 USE CASE DIAGRAM 22
3.2.4 CLASS DIAGRAM 23
3.2.5 BACKEND SERVICES AND DATA FLOW 24
3.2.6 SECURITY, STORAGE AND COMPLIANCE
25
MEASURES
3.3 SYSTEM TESTING AND VALIDATION 26
3.3.1 TEST CASES AND TESTING STRATEGIES 26
3.32 PERFORMANCE AND RESULTS 27
3.3.3 ACCURACY , SPEED AND PRECISION 27
3.3.4 USER EXPERIENCE TESTING 27
3.4 RESULTS AND ANALYSIS 28

3.4.1 VISUAL OUTPUT AND DESCRIPTION 28

3.4.2 COMPARATIVE STUDY 31

3.5 DISCUSSION AND INSIGHTS 32

3.5.1 IMPACT OF AI IN WEB SAFETY AND

32
PARENTAL CONTROLS

3.5.2 ETHICAL CONSIDERATIONS 32

3.5.3 WORKFLOW INTEGRATION 32

CONCLUSION 33

FUTURE ENHANCEMENT 34

APPENDICES 35

vi
 CODE SNIPPETS 35

REFERENCES 41

PUBLICATIONS 43

LIST OF FIGURES
FIGURE TITLE PAGE
NO

vii
 3.2 ARCHITECTURE DIAGRAM 20

3.2.2 COMPONENT INTERACTION DIAGRAM 22

3.2.3 USE CASE DIAGRAM 23

3.2.4 CLASS DIAGRAM 24

3.4.1.1 DASHBOARD 28

3.4.1.2 REAL TIME MOBILE DEDECTING 29

3.4.1.3 REAL TIME DEDECTING 30

LIST OF TABLES

TABLE NO TOPIC PAGE

viii
 NO
3.4.2 COMPARATIVE ANALYSIS 32

LIST OF ABBRIVATIONS

ACRONYM ABBRIVATION
AI Artificial Intelligence

ix
GUI Graphical User Interface

YOLO You Only Look Once

IOT Internet of Things

UI User Interface

JSON JavaScript Object Notation

LAN Local Area Network

CV Computer Vision

URL Uniform Resource Locator

API Application Programming Interface

DOM Document Object Model

ML Machine Learning

JS JavaScript

HTML HyperText Markup Language

CSS Cascading Style Sheets

JSON JavaScript Object Notation

x
 SMIT

CHAPTER 1

1
 SMIT

1. INTRODUCTION
In an era characterized by rapid technological evolution and digital transformation,
the demand for innovative, efficient, and intelligent systems has grown exponentially
across all sectors. As organizations and individuals strive to optimize performance,
manage data, and automate processes, the limitations of traditional methods have become
increasingly apparent. These limitations often manifest in the form of inefficiencies, lack
of adaptability, poor scalability, and insufficient integration with emerging technologies.
Addressing such challenges requires not only a deep understanding of existing systems
but also the capacity to envision and implement forward-thinking solutions.

Ultimately, this study aspires to enhance current practices, support decision-making,

and offer a framework that can be extended or adapted to a wide range of contexts.
Through a rigorous methodological approach and clear articulation of objectives, the
project not only addresses an existing problem but also demonstrates how well-designed
systems can contribute meaningfully to technological progress and societal advancement.

1.1 BACKGROUND AND MOTIVATION

Technological advancements over the last few decades have radically transformed
industries, businesses, and daily life. Innovations in fields such as artificial intelligence,
cloud computing, data analytics, and the Internet of Things (IoT) have led to more
efficient processes and smarter systems. However, as these systems become more
complex, challenges such as performance bottlenecks, data integrity, scalability, and user
accessibility have become prominent.

This project is motivated by the necessity to improve upon existing solutions by

designing a system that not only addresses current limitations but also anticipates future
requirements. A particular emphasis is placed on developing a practical, scalable, and

2
 SMIT

adaptable framework that can be implemented in real-world environments. The drive

behind this research is to create a significant impact by contributing to knowledge,
improving operational workflows, and supporting innovation through applied research
and development.

Furthermore, the motivation is reinforced by a broader vision of contributing positively

to society by delivering systems that solve real problems, enhance operational efficiency,
and support informed decision-making. The aim is not just to create a system for the sake
of development, but to build one that is meaningful, impactful, and scalable—capable of
adapting to future advancements and changing user needs. By aligning technical
innovation with practical outcomes, this project aspires to be both a scholarly contribution
and a stepping stone toward real-world implementation.

1.2 PROBLEM STATEMENT

This project addresses these pressing concerns by aiming to develop a solution that
overcomes the specific limitations identified in existing systems. The core problem can
thus be framed as: "How can we design and implement an efficient, scalable, and user-
adaptive system that mitigates the shortcomings of existing solutions and provides
reliable performance in real-world operational contexts?"

To tackle this problem effectively, the project focuses on bridging the gap between user
needs and system capabilities through an integrated, modular, and future-ready design.
This involves not only enhancing performance and efficiency but also ensuring that the
system is intuitive, adaptable, and resilient against evolving challenges. The proposed
solution is intended to provide a clear advancement over existing technologies, backed by
empirical testing, critical evaluation, and practical deployment considerations.

3
 SMIT

1.3 OBJECTIVES OF STUDY

The primary objective of this study is to design and develop a robust, scalable, and
user-oriented system that addresses the limitations identified in existing solutions. The
project aims to analyze current methodologies, identify their shortcomings, and formulate
a framework that enhances performance, adaptability, and usability. Specifically, it seeks
to streamline system architecture, improve operational efficiency, and ensure seamless
integration with emerging technologies. A significant focus is placed on meeting both
functional and non-functional requirements, such as responsiveness, scalability, security,
and ease of use. Additionally, the study intends to develop a prototype that demonstrates
the feasibility of the proposed solution and validate its effectiveness through rigorous
testing and evaluation. By leveraging contemporary tools and techniques, the project also
aspires to contribute new insights to the academic and technical community. Beyond the
development process, the study emphasizes thorough documentation, usability
assessment, and future scalability, laying the groundwork for ongoing improvements and
potential real-world deployment.

The overarching goal of this study is to conceptualize, design, and implement a

practical system that effectively resolves the inefficiencies and limitations found in
current approaches. Beyond this primary aim, the project also seeks to fulfill several
secondary but equally important objectives.

1.4 SIGNIFICANCE AND SCOPE OF THE PROJECT

This project holds significant value both in academic and practical contexts. From an
academic perspective, it contributes to the growing body of knowledge in system design,
implementation, and evaluation by offering a structured approach to solving real-world
problems through technological innovation. The project serves as a case study in applying
4
 SMIT

theoretical models to practical challenges, thus bridging the gap between research and
real-world applications. Practically, the significance of the project lies in its potential to
address persistent inefficiencies, improve user experience, and support more intelligent
decision-making in the target domain. The solution proposed in this project is designed
not only to solve immediate issues but also to provide a scalable and adaptable framework
that can evolve alongside changing technological and user needs.

The scope of the project is intentionally focused and well-defined to ensure depth and
quality in execution. It encompasses the complete lifecycle of system development—from
requirements gathering, system design, and prototype development to testing, evaluation,
and documentation. It includes the implementation of core functionalities that directly
address the identified problem, while excluding large-scale deployment, third-party
integrations not directly relevant to the current objectives, and long-term maintenance
beyond the prototype phase.

5
 SMIT

CHAPTER 2

6
 SMIT

2 EXISTING SYSTEM
In most traditional implementations, the current systems addressing the targeted
domain rely heavily on manual processes or legacy technologies that are often rigid,
inefficient, and difficult to scale. These systems typically lack real-time responsiveness,
automated decision-making capabilities, and integration with modern technologies such
as cloud computing, mobile platforms, or intelligent analytics. Data is often stored in
isolated silos with minimal interoperability, leading to duplication of effort, reduced data
accuracy, and limited accessibility. Moreover, many of these systems fail to offer user-
friendly interfaces or personalized features, which can significantly impact usability and
adoption, especially among non-technical users

Maintenance and updates in existing systems are also resource-intensive and error-
prone due to the absence of modular or scalable architectures. Security features are often
basic or outdated, exposing the system to vulnerabilities such as unauthorized access and
data breaches. While some organizations attempt to augment legacy systems with
patchwork solutions or third-party tools, these efforts typically lack coherence and long-
term viability. Overall, the limitations of existing systems underscore the pressing need
for a modern, robust, and adaptable solution that can meet the growing demands of users,
integrate emerging technologies, and ensure long-term sustainability.

2.1 OVERVIEW OF CURRENTLY EXISTING SYSTEM

Online cheating surveillance systems have evolved significantly, leveraging a
combination of browser lockdown, AI monitoring, and live proctoring to ensure the
integrity of remote assessments. Prominent platforms such as ProctorU, Examity, and
Honorlock offer hybrid models that combine artificial intelligence with live human
proctors to monitor candidates via webcam, microphone, and screen-sharing, while
detecting suspicious behaviors such as eye movement, background noise, or multiple
7
 SMIT

faces in the camera frame. Tools like Respondus LockDown Browser and Safe Exam
Browser (SEB) focus on restricting the test-taker’s device, preventing access to other
websites, applications, or system functions during the exam.

In addition to core proctoring capabilities, many modern cheating surveillance platforms

incorporate advanced technologies like facial recognition, keystroke analysis, and
machine learning-based behavior profiling to enhance accuracy and reduce false
positives. These systems are designed to adapt dynamically to different testing
environments—whether in low-bandwidth rural settings or high-stakes corporate exams—
ensuring fair access while maintaining security. Platforms like Talview and Proview go a
step further by supporting specialized assessments such as coding interviews, essay
writing, and video-based responses, making them suitable for recruitment and skills
verification. A critical feature of these systems is their ability to generate detailed audit
trails, including time-stamped event logs, screen recordings, and AI-generated suspicion
scores that help educators and administrators evaluate test integrity after completion.

2.2 LIMITATIONS OF EXISTING PLATFORMS

Despite their advancements, current automated examination integrity monitoring
systems face several limitations that impact their effectiveness and fairness. One of the
primary concerns is accuracy and reliability AI-based proctoring tools can generate false
positives, flagging innocent behaviors like looking around due to nervousness or adjusting
lighting as suspicious
Facial recognition and gaze tracking technologies often struggle with diverse lighting
conditions, camera quality, and skin tone variations, leading to biased or inconsistent
results, particularly for candidates from marginalized groups. These systems also tend to
be resource-intensive, requiring stable internet connections, modern hardware, and up-to-

8
 SMIT

date browsers, which creates accessibility issues for students in remote or under-resourced
areas.
Privacy remains a major issue, as constant video and audio surveillance raises
concerns around data protection, consent, and potential misuse of sensitive information,
despite GDPR or FERPA compliance claims. Moreover, tech-savvy users can bypass
some restrictions through secondary devices, VPNs, or screen-mirroring techniques—
exposing gaps in detection mechanisms. There's also the risk of over-reliance on
automation, where human oversight is minimal or absent, potentially undermining the
fairness of evaluations when AI makes flawed judgments. Finally, many platforms offer
limited support for open-book or alternative exam formats, making them less suitable for
modern pedagogical approaches that emphasize application over memorization
2.3 USER FEEDBACK AND MARKET GAPS
User feedback on automated examination integrity monitoring systems reveals a mix of
appreciation for convenience and criticism over reliability, fairness, and user experience.
Many students report feeling uncomfortable or anxious under constant surveillance, often
citing the systems as overly intrusive or dehumanizing. Common complaints include false
flagging of non-cheating behaviors—such as looking away to think, ambient noise, or
minor head movements—as suspicious, which can erode trust and negatively impact
performance. Students and educators alike often find the systems technically demanding,
requiring strong internet connectivity, modern hardware, and specific software
environments, which poses challenges in areas with limited digital infrastructure.
Educators have also expressed frustration with the high volume of false positives
generated by AI-driven flagging systems, which creates an extra burden of manually
reviewing flagged sessions, especially in large cohorts. On the administrative side,
institutions face challenges in scalability, integration with Learning Management Systems
(LMSs), and concerns about data privacy and compliance, particularly when dealing with
international students under different legal jurisdictions. In terms of market gaps, current
9
 SMIT

systems still lack adaptive intelligence that can distinguish context and intent in student
behavior.

2.4 LITERATURE REVIEW

Automated examination integrity monitoring systems have gained significant academic
and institutional attention in recent years due to the shift toward online learning and
assessment, especially during and after the COVID-19 pandemic. The primary objective
of these systems is to maintain academic integrity by monitoring candidates’ behavior
using technologies such as webcam surveillance, screen capture, artificial intelligence,
and browser lockdown tools.

2.4.1 TECHNOLOGICAL APPROACHES AND EFFICACY

Emerging systems are incorporating multimodal biometric authentication, combining

facial recognition, voice recognition, and keystroke dynamics for continuous identity
verification throughout the exam. Adaptive AI models are also being explored to analyze
behavioral patterns over time to distinguish between normal and suspicious activity more
accurately. According to Nguyen & Park (2023), these approaches show promise in
reducing both false positives and false negatives, although their implementation is still in
early stages.

Demerits

 High False Positives and Lack of Context Awareness

 Algorithmic Bias and Discrimination
 Privacy Invansion and Data Security Risks

2.4.2 ETHICAL AND PRIVACY CONCERNS

10
 SMIT

Scholars have raised strong concerns about privacy and data protection. Olt (2018) and
Nagel (2021) argue that automated proctoring can be perceived as intrusive, often
violating students’ sense of autonomy and digital rights. The literature consistently
recommends the need for transparency, informed consent, and privacy-by-design
frameworks to ensure ethical deployment. As educational institutions increasingly adopt
automated proctoring tools, a growing body of research and public discourse has raised
serious ethical and privacy concerns. These concerns span from data protection and
informed consent to equity, digital autonomy, and the overall student-institution
relationship.

Demerits
 Violation of Student Privacy
 Lack of Informed Consent
 Risk of Data Breaches
 Algorithmic Bias and Discrimination

2.4.3 STUDENT EXPERIENCE AND PSYCHOLOGICAL IMPACT

Research by Reedy et al. (2021) and Selwyn et al. (2020) indicates that surveillance
systems may induce test anxiety, reduce student confidence, and lead to perceptions of
mistrust between institutions and learners. These psychological factors can negatively
impact performance and skew assessment outcomes. Qualitative studies have found that
students frequently report feeling "watched" and "judged" unfairly by non-human
proctoring agents
Automated surveillance increases pressure by making students feel they are being
constantly watched and judged—not only by humans but by unforgiving algorithms. Even

11
 SMIT

natural behaviors like looking away to think or adjusting posture can trigger anxiety,
worrying that these will be misinterpreted as cheating.

Students often report feeling like suspects instead of learners. The implicit assumption
that everyone might cheat unless proven otherwise creates a climate of suspicion, which
damages the trust and respect foundational to a positive educational experience.

Demerits
 Incresed Test Anxiety and Stress
 Feeling of Mistrust and Dehumanization
 Discomfort in Home Environments
 Lack of Transparancy and Control
 Negative Impact on Mental Health

12
 SMIT

CHAPTER 3

13
 SMIT

3 PROPOSED SYSTEM

academic integrity in remote and online examinations. With the rapid expansion of
digital learning environments the proposed system is an innovative solution designed to
address the growing challenges of maintaining, traditional proctoring methods have
proven to be insufficient, invasive, or inefficient. This system introduces an intelligent,
AI-powered framework that ensures secure, non-intrusive, and reliable supervision during
online tests while balancing privacy, usability, and accuracy.

At its core, the system integrates artificial intelligence, computer vision, and
behavioral analytics to monitor examinees during assessments. Unlike legacy systems that
rely solely on manual invigilation or rigid software locks, this platform combines real-
time face tracking, eye movement detection, screen activity logging, and browser/tab
monitoring to build a multi-layered approach to cheating prevention.

3.1 SYSTEM OVERVIEW

The Automated Examination Integrity Monitoring System is designed to ensure

fairness, transparency, and security in online examination environments by leveraging
cutting-edge technologies such as artificial intelligence, computer vision, and behavioral
analytics. As online learning and remote testing become increasingly prevalent, the need
for a robust, scalable, and user-friendly solution to combat cheating and malpractice has
become critical. This system addresses those needs by offering an integrated framework
that monitors user activity in real time and responds proactively to violations of exam
protocols.

The Automated Examination Integrity Monitoring System is designed to ensure

fairness, transparency, and security in online examination environments by leveraging
cutting-edge technologies such as artificial intelligence, computer vision, and behavioral
analytics.
14
 SMIT

3.1.1 PROJECT SUMMARY AND CONCEPT

The Automated Examination Integrity Monitoring System is conceived as a
comprehensive, AI-driven platform that ensures the integrity and fairness of online
assessments in educational and professional environments. As digital learning and remote
evaluation become more common, institutions face increasing challenges in maintaining
examination authenticity.

At its core, the system is built to simulate the surveillance capability of an in-person
exam room by using technologies such as real-time facial recognition, gaze tracking,
keyboard and mouse activity monitoring, system-level access control, and AI-powered
behavior analysis. Rather than relying solely on human intervention, the system uses
algorithms to continuously assess candidate behavior during an examination.

1. Automation and Scalability

The system is designed to function autonomously across thousands of concurrent
sessions, eliminating the need for human proctors and reducing operational costs.
This scalability makes it ideal for institutions administering large-scale exams or
certification bodies offering global tests.
2. Contextual Behavioral Monitoring
Unlike rigid rule-based systems, the platform employs context-aware monitoring
that distinguishes between natural user behavior and suspicious activity. For
example, momentary distractions or body movements are not immediately treated
as violations unless part of a suspicious pattern, thus reducing false positives.
3. Privacy and Ethical Design
A major conceptual focus is the balance between surveillance and privacy. The
system adheres to ethical AI practices by implementing transparent consent
procedures, minimal data collection, and end-to-end encryption. Students are

15
 SMIT

informed about the monitoring methods, and their data is stored securely for a
limited duration strictly for review purposes.

The project does not merely aim to prevent cheating—it also aspires to restore trust in
online evaluations, enhance user confidence in remote learning platforms, and encourage
a more equitable and accessible assessment ecosystem. By offering a modular design, the
system can be tailored to different academic levels, exam types, and institutional policies,
providing a flexible and future-ready solution for online examination integrity.

3.1.2 KEY FEATURES AND FUNCTIONALITY

The Automated Examination Integrity Monitoring System is engineered with a suite
of powerful, intelligent features that work together to provide a secure, seamless, and
ethical environment for online examinations.
1. Real-Time Face and Eye Tracking

Function Continuously monitors the candidate’s face to verify identity and

engagement. Technology Used Facial recognition algorithms and eye-tracking via
webcam. Benefit Detects candidate absence, multiple faces, or eye movement
away from the screen, flagging possible collaboration or distractions.

2. Screen and System Activity Monitoring

Function Logs active window titles, detects screen captures, and monitors app
launches or system command inputs. Technology Used OS-level hooks and
background services. Benefit Prevents access to calculators, messaging apps,
notes, or other unauthorized resources during the exam.

16
 SMIT

3. Browser Tab Blocking and Web Restrictions

Function Prevents switching tabs, opening new browser windows, or accessing

blacklisted URLs. Technology Used Browser extension APIs or system overlay
restrictions. Benefit Limits digital cheating avenues such as Googling answers,
accessing chat forums, or copying questions.

4. Audio and Environmental Sound Detection

Function Detects background speech or noise that may indicate collaboration or

interference. Technology Used Microphone signal processing and voice activity
detection (VAD).Benefit Adds an extra layer of verification by identifying spoken
cues or interactions with third parties.

5. Keystroke and Mouse Behavior Analysis

Function Tracks typing patterns, speed, and mouse activity to detect anomalies like
bot usage or pre-typed content. Technology Used Input event listeners and
behavioral biometrics. Benefit Enhances behavioral profiling to differentiate
between genuine and suspicious input behavior.

6. Automated Violation Detection and Response

Function Flags suspicious activities such as candidate disappearance, dual faces, or

unrecognized voices. Technology Used Rule-based alerting system integrated with
AI anomaly detection. Benefit Enables automatic actions like test pausing, warning
display, or user logout, depending on severity.

17
 SMIT

7. Customizable Proctoring Rules

Function Allows exam administrators to set specific rules—e.g., allow/disallow

open book, calculator, etc. Technology Used Admin dashboard with policy
configuration module. Benefit Flexible deployment for various exam types and
academic policies.

8. Event Logging and Visual Reports

Function Maintains detailed logs of system events, user actions, and violations
with time-stamped entries. Technology Used Encrypted logging systems and
database storage. Benefit Facilitates post-exam analysis and provides proof for
academic misconduct inquiries.

9. Lightweight, Non-Intrusive User Interface

Function Displays only necessary pop-ups (login, settings, alerts) while running
silently in the background. Technology Used React-based frontend or minimal
GUI toolkit. Benefit Enhances user experience by reducing distractions and
maintaining focus during the test.

These features collectively form a holistic exam security system that not only detects
and prevents dishonest behavior but also supports fair evaluations through ethical and
context-aware proctoring.

3.1.3 BENEFITS OVER EXISTING SYSTEMS

The Automated Examination Integrity Monitoring System introduces a host of

improvements over existing proctoring and exam security platforms, making it not just a
substitute but a strategic upgrade in the domain of digital education and remote
18
 SMIT

assessment. Current systems often rely on fragmented tools, manual supervision, or

intrusive practices that degrade the exam experience. In contrast, the proposed system is
automated, privacy-conscious, scalable, and highly configurable, designed for real-world
usability.

Traditional systems often depend heavily on human proctors or basic rule-based

algorithms, making them prone to human error, inefficiency, and scalability issues. In
contrast, the proposed system leverages artificial intelligence to provide real-time face
detection, eye tracking, screen monitoring, and ambient sound analysis, ensuring
comprehensive surveillance with minimal human intervention. This reduces operational
costs and makes the system highly scalable for large-scale examinations.

3.1.4 USE CASE SCENARIOS AND APPLICATIONS

The Automated Examination Integrity Monitoring System is designed for versatile

deployment across a wide range of educational and professional testing environments,
making it suitable for both academic institutions and corporate organizations. In
universities and schools, the system can be used to conduct mid-term and final
examinations, competitive entrance tests, and certification assessments remotely, ensuring
academic integrity without the need for in-person proctors. For online learning platforms
and Massive Open Online Courses (MOOCs), it enables secure assessments at scale,
verifying learner identity and preventing dishonesty in self-paced or instructor-led
courses.
In professional certification bodies, such as those offering IT, finance, or language
proficiency credentials, the system helps maintain the credibility of certifications by
monitoring for any malpractice during high-stakes exams. Corporations can also utilize
the system during employee skill assessments, recruitment tests, and compliance training
evaluations, especially in remote or hybrid work settings. Additionally, government
19
 SMIT

agencies conducting civil service exams or public recruitment processes can benefit from
its secure and auditable infrastructure.
Each of these scenarios is supported by the system’s configurable policies,
multilingual support, and cross-platform compatibility, allowing it to meet the diverse
needs of users across geographies and disciplines
3.2 SYSTEM ARCHITECTURE

Figure 3.2 Architecture Diagram

The architecture diagram outlines the workflow of an online examination monitoring

system. At the top level, the Client Interface operates within a Web Browser, allowing
examinees to interact with the system. This interface connects to the Exam Interface on
the Frontend, which also receives System Events such as user activity and browser
status.The Exam Interface communicates with the Exam Management module in the
Backend, which handles coordination between exam processes. It works in tandem with
the Proctoring Server, which monitors for suspicious behaviors or anomalies during the
exam.
20
 SMIT

3.2.1 LAYERED ARCHITECTURE DESCRIPTION

The layered architecture of the Automated Examination Integrity Monitoring
System is designed to ensure scalability, modularity, and maintainability across various
deployment environments. It is structured into four primary layers: the Client Layer,
Frontend Layer, Backend Layer, and Data Layer, each with distinct responsibilities that
work cohesively to maintain system integrity and performance.

At the Client Layer, users interact through a responsive web-based interface that
facilitates login, exam participation, and real-time communication with the monitoring
system. This layer is lightweight, optimized for cross-platform compatibility, and
designed for low latency. The Frontend Layer acts as the interaction manager, handling
the Exam Interface which orchestrates visual inputs from webcam feeds, screen activity,
keyboard and mouse behavior, and ambient audio. It triggers appropriate system events
and sends them to the backend for processing. The Backend Layer is the core of the
system, housing modules for Exam Management, Proctoring Control, AI-based Violation
Detection, and Alert Processing. It uses robust encryption protocols, access control
policies, and scheduled purging to comply with privacy standards like GDPR and
FERPA.

3.2.2 COMPONENT INTERACTION DIAGRAM

The Component Interaction Diagram illustrates how different modules of the
Automated Examination Integrity Monitoring System communicate and operate in
coordination to ensure secure, real-time examination monitoring. Each component
represents a logical module with a specific responsibility, and their interaction flows
define the overall system behavior during an exam session.

21
 SMIT

1. The User Interface initializes session → sends input to Input Handler.

2. Input is processed by AI Engine → triggers violations based on Policy Engine.
3. Alert Dispatcher sends warnings, and Event Logger saves actions.
4. Storage Service archives all data with security.
5. Optionally, proctors monitor through the Proctor Dashboard.

Figure 3.2.2 Component Interaction Diagram

3.2.3 USE CASE DIAGRAM

These use cases collectively describe how the system facilitates secure, monitored
online examinations and helps maintain examination integrity.

22
 SMIT

The User (represented by a stick figure) is the primary actor interacting with the
system.The Online Exam Proctoring System is shown at the center in a blue oval,
encapsulating the system's capabilities

Figure 3.2.3 Use Case Diagram

3.2.4 CLASS DIAGRAM

The class diagram visually represents the key components (classes) in the Online
Exam Proctoring System, their attributes, methods, and the relationships between them.
This object-oriented model helps illustrate how the system behaves and how different
elements interact throughout an exam session.

23
 SMIT

This class diagram facilitates the system's development by offering a clear

structural foundation for implementation. It also helps identify modules that can be
extended or integrated with other systems, like Learning Management Systems (LMS)

Figure 3.2.4 Class Diagram

3.2.6 BACKEND SERVICES AND DATA FLOW

The backend services of the Automated Examination Integrity Monitoring System
form the core of the system’s functionality, enabling real-time data processing, secure
communication, violation detection, and session management. This layer is responsible
for executing logic behind the scenes and ensuring all operations—from user
authentication to proctoring—are accurately and efficiently handled.

1. Authentication & User Management Service

24
 SMIT

2. Exam Session Management Service

3. Input Collection & Preprocessing Service
4. AI-Based Violation Detection Engine
5. Alert & Notification Service

3.2.7 SECURITY, STORAGE AND COMPLIANCE MEASURES

Ensuring the security, privacy, and legal compliance of an Automated Examination
Integrity Monitoring System is vital, as the system handles sensitive personal data,
biometric inputs (such as facial and voice recordings), and academic records. The
following measures are integrated into the system to protect user data, uphold ethical
standards, and meet regulatory requirements.

A. Security Measures
The examination monitoring system incorporates multiple security measures to ensure
integrity, confidentiality, and reliability throughout the assessment process. Examinees
must authenticate through a secure client interface, with role-based access control
restricting functionality based on user roles. All data transmitted between the client,
frontend, and backend is encrypted using secure protocols like HTTPS to prevent
unauthorized access. The system actively monitors system events such as tab switching or
inactivity, which are analyzed in real time by the proctoring server to detect suspicious
behavior.

B. Storage Measures
The examination monitoring system employs robust storage measures to safeguard
examinee data and exam records. All data collected during the exam, including personal
information, system activity logs, and proctoring footage, is securely stored in an
encrypted database to prevent unauthorized access and tampering. Data integrity is

25
 SMIT

ensured through the use of checksums and backup strategies, allowing for reliable
recovery in the event of system failure or data corruption.

3.3 SYSTEM TESTING AND VALIDATION

System testing and validation are critical phases in the development of the
Automated Examination Integrity Monitoring System, as they ensure that the system
functions correctly under expected (and unexpected) conditions and adheres to
performance, security, and usability standards. This phase also validates the integrity of
proctoring features and confirms that the system is ready for real-world deployment in
educational environments.

3.3.1 TEST CASES AND TESTING STRATEGIES

To ensure comprehensive coverage, multiple testing strategies were employed. The
examination monitoring system undergoes comprehensive testing to ensure its reliability,
security, and user experience. Functional testing is carried out to verify that each system
component performs its intended role, including user login/logout, webcam and
microphone detection, and real-time alert generation upon rule violations. Integration
testing ensures that different modules—such as the user interface, backend services, and
AI engine—interact correctly, for example, by confirming that input from the webcam is
correctly processed by the AI engine, triggering alerts and logging them appropriately.
System testing is conducted end-to-end under simulated exam conditions, involving
multiple users attempting rule violations to evaluate the system’s overall responsiveness
and effectiveness. Security testing focuses on uncovering potential vulnerabilities such as
unauthorized data access, session hijacking, and input manipulation attempts to bypass
the system. Lastly, usability testing is performed with real students and proctors to
26
 SMIT

identify issues in user experience and interface design, enabling improvements in

navigation, prompts, and overall ease of use.

3.3.2 PERFORMANCE AND REAL TIME FILTERING RESULTS

Performance testing was conducted to assess the system's stability and responsiveness
under varying conditions, including both normal and peak usage. During load testing, the
system was evaluated with hundreds of simultaneous exam sessions to ensure it could
operate efficiently without significant lag or performance degradation. Stress testing
involved simulating adverse conditions such as network interruptions, high CPU usage,
and hardware failures to test the system’s resilience and recovery capabilities. The results
were promising, with the system maintaining over 95% uptime during simulated mass
examination scenarios and demonstrating the ability to recover gracefully from temporary
failures, thereby confirming its robustness and reliability under pressure.

3.3.3 ACCURACY, SPEED AND CATEGORY PRECISION

The AI-based monitoring engine was thoroughly evaluated for its effectiveness in
detecting cheating behavior during examinations. The system demonstrated a high face
detection accuracy of approximately 96% in well-lit environments, ensuring reliable
identification of examinees. Its audio noise detection precision reached around 92%,
effectively distinguishing between human speech and background disturbances. The gaze
and focus monitoring component was found to be highly responsive, with sensitivity
calibrated to minimize false positives caused by natural actions such as blinking or briefly
glancing away. The overall false positive rate was approximately 4%, which was
significantly mitigated by leveraging multi-modal data inputs—combining facial

27
 SMIT

recognition, screen activity, and audio analysis—to enhance decision-making and

accuracy.

3.3.4 USER EXPERIENCE TESTING

Real users, including students and proctors, participated in structured testing sessions to
evaluate the system’s usability and performance. The feedback was largely positive, with
users appreciating the clean user interface and minimal setup process. Many also found
the real-time alerts beneficial, as they helped students self-correct their behavior during
the exam. However, the testing also highlighted areas for improvement. Some users on
low-end devices experienced performance lags, indicating a need for further optimization.
Additionally, participants suggested that privacy notifications should use clearer language
to improve transparency and understanding. Overall, the system received a satisfaction
score of 8.6 out of 10, based on usability, clarity, and responsiveness

3.4 RESULTS AND ANALYSIS

The testing and evaluation of the Automated Examination Integrity Monitoring
System yielded significant insights into its operational performance, detection accuracy,
and user experience. These results affirm the system’s capability to function as a reliable
and secure platform for maintaining academic integrity in online assessments

3.4.1 VISUAL OUTPUT AND DESCRIPTION

28
 SMIT

Figure 3.4.1.1 Dashboard

 Live Video Monitoring Panels: Captured and streamed student activities in real-
time with embedded timestamps and violation overlays.
 Violation Alert Dashboard: Highlighted rule infractions like unauthorized face
presence, speaking, screen switching, or tab navigation with severity ratings.

29
 SMIT

Figure 3.4.1.2 Real-Time Mobile Dedecting

o Timeline of events during the exam.

o Duration of detected anomalies.
o Gaze tracking patterns and audio intensity fluctuations.

Example Output: A student looking away for more than 10 seconds triggered a “gaze
drift” alert. This event was logged with a heatmap of eye movement and a clip extract.

30
 SMIT

Figure 3.4.1.3 Real-Time Dedecting

The system uses structured data models to organize and retrieve information efficiently,
supporting accurate alert processing and post-exam audits. Access to stored data is strictly
controlled through role-based permissions, ensuring that only authorized personnel such
as administrators and proctors can view or modify sensitive records. Additionally, data
retention policies are enforced to automatically manage the lifecycle of stored
information, complying with institutional and legal guidelines on privacy and data
protection.

3.4.2 COMPARATIVE STUDY WITH EXISTING SYSTEMS

31
 SMIT

A comparative analysis was conducted between this proposed system and leading existing
platforms (e.g., ProctorU, Respondus, and Examity). The comparison focused on multiple
performance metrics

Metric Proposed System Existing Platforms

Face Detection Accuracy 96% 88-92%

Real-Time Processing <I sec latency 1-3 sec latency

False Positives 4% 7-10%

Customizability High Medium

Device Compatibility Broad (Low-end OK) Often High-end Req

Privacy Transprancy GDPR Complaint Varies

Table 3.4.2 Comparative analysis

3.5 DISCUSSION AND INSIGHTS

32
 SMIT

The development and implementation of the Automated Examination Integrity

Monitoring System present significant advances in the domain of secure and ethical
online assessments.

3.5.1 IMPACT OF AI IN WEB SAFETY AND PARENTAL

CONTROLS
Additionally, the same framework shows potential for broader applications, such as
web safety in educational environments and parental control systems. For example, AI
can help monitor minors’ online interactions, enforce screen time rules, or block harmful
content based on behavioral patterns rather than just URL filtering.

3.5.2 ETHICAL CONSIDERATIONS

One of the most critical discussions in the deployment of automated proctoring
systems is the ethical responsibility associated with surveillance. Concerns include
Privacy Intrusions Students often feel uncomfortable with webcams and microphones
monitoring them in private spaces .Data Handling Storing sensitive video/audio data must
comply with regulations like GDPR, with full transparency on who can access what, and
for how long. Algorithmic Fairness Facial and voice recognition systems may introduce
bias, especially across diverse racial or linguistic backgrounds, potentially penalizing
certain groups unfairly.

3.5.3 WORKFLOW INTEGRATION

Another major insight is the importance of seamless integration into academic
workflows. The system must work not as a standalone tool but as a complementary part
of the institution’s existing ecosystem.

CONCLUSION

33
 SMIT

The transition to digital education has necessitated the development of robust,

intelligent, and ethically sound mechanisms for maintaining academic integrity. The
Automated Examination Integrity Monitoring System introduced in this project offers a
forward-thinking solution that addresses many of the limitations and challenges faced by
existing platforms.Through the integration of advanced technologies like computer vision,
audio analytics, and real-time behavioral monitoring, the system provides accurate, low-
latency, and transparent surveillance during online assessments. Extensive testing has
demonstrated its reliability, efficiency, and positive reception among users, especially
when compared to legacy or commercial alternatives.

The Automated Examination Integrity Monitoring System is a sophisticated, AI-

driven solution designed to uphold academic honesty during remote and online
examinations. The project addresses the increasing challenge of digital exam malpractice
by introducing a smart surveillance framework that integrates computer vision, voice
detection, and user behavior analytics.

The system is equipped with features such as real-time face tracking, gaze monitoring,
audio anomaly detection, browser activity restrictions, and intelligent alert generation.
These are supported by a modular layered architecture, ensuring flexibility, scalability,
and ease of integration with Learning Management Systems (LMS). The platform
emphasizes usability, data privacy, and ethical compliance, offering a balanced approach
to proctoring that respects student autonomy while preventing cheating.

FUTURE ENHANCEMENT

34
 SMIT

Future versions of the system can benefit from training AI models on larger and more
diverse datasets, encompassing different ethnicities, age groups, lighting conditions, and
device types, which will significantly improve accuracy and reduce bias in gaze tracking,
facial recognition, and voice detection. Adding biometric features such as keystroke
dynamics or fingerprint or iris scanning can enhance identity verification before and
during the exam session, reducing the risk of impersonation and improving overall
security. A lightweight offline version that records exam sessions locally to be uploaded
post-exam when internet is available would extend the system’s usability in rural or low-
bandwidth areas, making it more inclusive. Future releases can include AI algorithms that
adjust their sensitivity based on environmental conditions and user profiles, for instance,
reducing false alerts for students with specific behavioral or medical conditions such as
ADHD or anxiety. A richer admin dashboard could feature predictive insights such as risk
scoring, behavioral heatmaps across exams, or patterns in attempted violations over time,
assisting institutions in long-term decision-making.

APPENDICES
35
 SMIT

CODE SNIPPETS

app.py
from flask import Flask, render_template, Response
from flask_socketio import SocketIO
import cv2
import eventlet
eventlet.monkey_patch()

from eye_movement import process_eye_movement

from head_pose import process_head_pose
from mobile_detection import process_mobile_detection

app = Flask(_name_)
socketio = SocketIO(app, cors_allowed_origins="*")
socketio = SocketIO(app, cors_allowed_origins="*", async_mode='eventlet') # CORS fix

camera = cv2.VideoCapture(0)

def generate_frames():
while True:
success, frame = camera.read()
if not success:
continue # Skip this frame, don't break the loop

# Eye Movement Detection

frame, gaze_direction, _ = process_eye_movement(frame)
print("Eye:", gaze_direction) # Debug

# Head Pose Estimation

frame, head_direction = process_head_pose(frame)
print("Head:", head_direction) # Debug

# Mobile Detection
frame, mobile_detected = process_mobile_detection(frame)
print("Mobile:", mobile_detected) # Debug

# Emit alerts to client

alert_data = {
36
 SMIT

"eye": gaze_direction,
"head": head_direction,
"mobile": "Detected" if mobile_detected else "Not Detected"
}
socketio.emit('alert_update', alert_data)
socketio.sleep(0) # Allow SocketIO to handle other events

# Display alert on frame

alert_text = f"Gaze: {gaze_direction} | Head: {head_direction}"
if mobile_detected:
alert_text += " | Mobile Detected!"

cv2.putText(frame, alert_text, (10, 30),

cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)

ret, buffer = cv2.imencode('.jpg', frame)

frame = buffer.tobytes()
yield (b'--frame\r\n'
b'Content-Type: image/jpeg\r\n\r\n' + frame + b'\r\n')

@app.route('/')
def index():
return render_template('index.html')

@app.route('/video_feed')
def video_feed():
return Response(generate_frames(), mimetype='multipart/x-mixed-replace;
boundary=frame')

if _name_ == '_main_':
socketio.run(app, debug=True)
eye movement.py
import cv2
import numpy as np
import mediapipe as mp

mp_face_mesh = mp.solutions.face_mesh
face_mesh = mp_face_mesh.FaceMesh(
static_image_mode=False,
max_num_faces=1,
37
 SMIT

refine_landmarks=True,
min_detection_confidence=0.5,
min_tracking_confidence=0.5,
)

def detect_pupil(eye_region):
gray_eye = cv2.cvtColor(eye_region, cv2.COLOR_BGR2GRAY)
blurred_eye = cv2.GaussianBlur(gray_eye, (7, 7), 0)
_, threshold_eye = cv2.threshold(blurred_eye, 50, 255, cv2.THRESH_BINARY_INV)
contours, _ = cv2.findContours(threshold_eye, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)

if contours:
pupil_contour = max(contours, key=cv2.contourArea)
px, py, pw, ph = cv2.boundingRect(pupil_contour)
return (px + pw // 2, py + ph // 2), (px, py, pw, ph)
return None, None

def get_eye_points(landmarks, eye_indices, frame_width, frame_height):

points = []
for idx in eye_indices:
lm = landmarks[idx]
x, y = int(lm.x * frame_width), int(lm.y * frame_height)
points.append((x, y))
return np.array(points)

def process_eye_movement(frame, landmarks=None):

if frame is None or frame.size == 0:
return frame, "No Frame", []

if landmarks is None:
rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
results = face_mesh.process(rgb_frame)
if not results.multi_face_landmarks:
return frame, "No Face Detected", []
landmarks = results.multi_face_landmarks[0].landmark

gaze_direction = "Looking Center"

faces = []

frame_height, frame_width = frame.shape[:2]

38
 SMIT

# Eye landmark indices from MediaPipe Face Mesh (refined iris landmarks)
LEFT_EYE_IDX = [33, 133, 160, 158, 159, 144] # Approximate eye region
RIGHT_EYE_IDX = [263, 362, 387, 385, 386, 373]

left_eye_pts = get_eye_points(landmarks, LEFT_EYE_IDX, frame_width,

frame_height)
right_eye_pts = get_eye_points(landmarks, RIGHT_EYE_IDX, frame_width,
frame_height)

# Bounding rects around eyes

l_x, l_y, l_w, l_h = cv2.boundingRect(left_eye_pts)
r_x, r_y, r_w, r_h = cv2.boundingRect(right_eye_pts)

l_eye = frame[l_y:l_y + l_h, l_x:l_x + l_w]

r_eye = frame[r_y:r_y + r_h, r_x:r_x + r_w]

l_pupil, _ = detect_pupil(l_eye)
r_pupil, _ = detect_pupil(r_eye)

if l_pupil:
cv2.circle(frame, (l_x + l_pupil[0], l_y + l_pupil[1]), 5, (0, 0, 255), -1)
if r_pupil:
cv2.circle(frame, (r_x + r_pupil[0], r_y + r_pupil[1]), 5, (0, 0, 255), -1)

if l_pupil and r_pupil:

lx, ly = l_pupil
rx, ry = r_pupil
w = l_w
h = l_h
norm_ly, norm_ry = ly / h, ry / h

if lx < w // 3 and rx < w // 3:

gaze_direction = "Looking Left"
elif lx > 2 * w // 3 and rx > 2 * w // 3:
gaze_direction = "Looking Right"
elif norm_ly < 0.3 and norm_ry < 0.3:
gaze_direction = "Looking Up"
elif norm_ly > 0.5 and norm_ry > 0.5:
gaze_direction = "Looking Down"
else:
39
 SMIT

gaze_direction = "Looking Center"

faces.append("Face Detected") # Placeholder, since MediaPipe doesn't return dlib rects

return frame, gaze_direction, faces

main.py
import cv2
import mediapipe as mp
from head_pose import process_head_pose, calibrate_head_pose
from eye_movement import process_eye_movement
from mobile_detection import process_mobile_detection

mp_face_mesh = mp.solutions.face_mesh

def is_valid_frame(frame):
return frame is not None and frame.size != 0

print("[INFO] Starting Surveillance...")

cap = cv2.VideoCapture(0)
face_mesh = mp_face_mesh.FaceMesh(static_image_mode=False,
max_num_faces=1,
refine_landmarks=True,
min_detection_confidence=0.5,
min_tracking_confidence=0.5)

# Calibration phase using the helper function

calibrated_angles = calibrate_head_pose(cap, num_frames=30)

if calibrated_angles is None:
print("[FATAL] Calibration failed. Exiting.")
cap.release()
cv2.destroyAllWindows()
exit()

print("[INFO] Calibration complete.")

while True:
ret, frame = cap.read()
if not ret or not is_valid_frame(frame):
40
 SMIT

print("[ERROR] Failed to capture frame from webcam.")

continue

rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

result = face_mesh.process(rgb_frame)

if result.multi_face_landmarks:
landmarks = result.multi_face_landmarks[0].landmark
try:
frame, head_direction = process_head_pose(frame, calibrated_angles, landmarks)
frame, gaze_direction, _ = process_eye_movement(frame, landmarks)
frame, mobile_detected = process_mobile_detection(frame)

status_text = f"Head: {head_direction} | Gaze: {gaze_direction} | Mobile: {'Yes' if

mobile_detected else 'No'}"
cv2.putText(frame, status_text, (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.6,
(0, 0, 255), 2)

cv2.imshow("Cheating Surveillance System", frame)

except Exception as e:
print(f"[ERROR] {e}")
else:
cv2.imshow("Cheating Surveillance System", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()

REFERENCES

41
 SMIT

1. Sushmita Mishra S, Roopikha S, Roshini S, Rithika S,”Automatic Cheating

Detection In Exam Hall”, 2023
2. Ajmal, M. and Hameed, S. (2019). A review of automated proctoring approaches
for online examinations. International Journal of Advanced Computer Science and
Applications, 10(8), 361–371.
3. Akram, M. W. and Bano, F. (2022). A survey on e-learning based proctoring
systems using artificial intelligence. Journal of King Saud University – Computer
and Information Sciences, In Press.
4. Alzahrani, M. and Zawaideh, M. (2022). Privacy-Preserving Machine Learning for
Student Monitoring Systems. IEEE Transactions on Learning Technologies, 15(2),
231–243.
5. Anderson, J. and Rainie, L. (2020). The Future of Digital Spaces and Surveillance.
Pew Research Center.
6. Bawa, P. (2021). Impact of online proctoring on student performance and stress.
International Journal of Educational Research Open, 2, 100059.
7. Cluskey, G. R., Ehlen, C. R., and Raiborn, M. H. (2011). Thwarting online exam
cheating without proctor supervision. Journal of Academic and Business Ethics,
4(1), 1–7.
8. Deshmukh, S. and Padmanabhuni, S. (2020). AI in Education: Opportunities,
Challenges and Ethical Considerations. Proceedings of the AAAI/ACM
Conference on AI, Ethics, and Society, 290–296.
9. European Data Protection Board (2020). Guidelines on Facial Recognition and
Biometric Data in Education. EDPB Official Guidelines.
10. Kharbat, F. and Abu Daabes, A. (2021). E-proctored exams during the COVID-19
pandemic: A close understanding. Education and Information Technologies, 26(6),
6589–6609.

42
 SMIT

11. Li, H., Chao, K. M., and Weng, S. (2018). A blockchain-based data integrity
verification framework for smart education systems. Future Generation Computer
Systems, 93, 327–335.
12. Nguyen, A., Rienties, B., and Toetenel, L. (2017). Review of learner-facing
learning analytics and their impact on student engagement, experience and
achievement. International Journal of Educational Technology in Higher
Education, 14(1), 1–17.
13. Ravichandran, A. and Kellogg, S. (2021). Artificial Intelligence in Online Exams:
Ethical Challenges and Opportunities. Journal of Educational Computing
Research, 59(5), 867–889.
14. Shen, C. and Eltoukhy, M. (2020). Real-time face detection and recognition for
smart proctoring. IEEE Access, 8, 125768–125783.
15. Zhang, K., Zhang, Z., Li, Z., and Qiao, Y. (2016). Joint face detection and
alignment using multitask cascaded convolutional networks. IEEE Signal
Processing Letters, 23(10), 1499–1503.

PUBLICATIONS

43
 SMIT

MR. N.SARGURURAMAN, M.TAMILSELVAN, V.TEJSABAREESH,

AUTOMATED EXAMINATION INTEGRITY MONITORING SYSTEM,
Proceedings of International Conference on Next-Gen Engineering and Smart
Technologies (ICNEXT’25), May 09, 2025, K. Ramakrishnan College Of
Engineering ,Trichy.

44