Title of the project

A project report submitted in the partial fulfilment of the requirements for the
award of the Degree
of

MASTER OF COMPUTER APPLICATIONS

Submitted By
Name of the student (Roll number)

Under the guidance of

Name of the Guide

Designation
Department

Department of Computer Applications

School of Computing and Informatics

Vadlamudi, Guntur, Andhra Pradesh - 522 213, India.

MAY - 2025
 CERTIFICATE

This is to certify that the project report entitled “Title of the project”
being submitted by Student name(Roll no) in partial fulfilment of the
requirements for the award of Master of Computer Applications, in the
Department of Computer Applications, Vignan’s Foundation for Science,
Technology and Research, (Deemed to be University), Vadlamudi, Guntur
District, Andhra Pradesh, India, is a bonafide work carried out by him under our
guidance and supervision.

Dr. Srikanthyadav.M Dr. SrikanthYadav. M Dr. K. Santhi Sri

Project Guide Project Coordinator HOD CA

External Examiner

ii
 DECLARATION

I hereby declare that my project work described in this titled “Automated

Diabetic Retinopathy Grading Using Graph Neural Networks for Enhanced
Feature Relationship Analysis in Retinal Images” which is being submitted by me
for the partial fulfilment of the requirements for the award of Master of Computer
Applications, in the Department of Computer Applications, Vignan’s Foundation for
Science, Technology and Research, (Deemed to be University), Vadlamudi, Guntur
District, Andhra Pradesh, and the result of investigations are carried out by me under
the guidance of Dr. srikanthyadav.M, Assistant Professor, Department of Computer
Applications.

Place : Vadlamudi K. Srinivasarao (231FD01028)

Date :

iii
 ACKNOWLEDGEMENT

My sincere thanks to various personalities responsible for the project's

completion. I am extremely thankful to my beloved Vice Chancellor, Dr. P.
Nagabhushan, who took a keen interest in me and every effort throughout this
course.

I express my deep-felt gratitude to my School of Computing & Informatics,

Joint Dean Dr. N. Veeranjaneyulu, Head of the Department Dr. K. Santhi Sri, and
Dr. SrikanthYadav. M coordinator of the project for extending their encouragement.
Their profound knowledge and willingness have been a constant source of inspiration
for me throughout the project work.

I wish to express my sincere, deep sense of gratitude to my guide, Dr.

Srikanthyadav.M significant suggestions and help in every respect to accomplish the
project work. His persisting encouragement, everlasting patience, and keen interest in
discussions have benefited me to be extent that cannot be spanned by words to my
college management for providing excellent lab facilities for the completion of the
project within my campus.

Finally, I extend my wholehearted gratitude to all my faculty members of the

Department of Information Technology & Computer Applications who helped me in
my academics throughout the course.

With Sincere Regards,

K. Srinivasarao (231FD01028)

iv
 PROJECT EXPERIENCE INFORMATION

1. Title of the Project :

__________Research_________________
2. Nature of the project :
(Design / Fabrication / Research)

3. Project Category : ____________Core_______________________

(Interdepartmental / Societal Centric/ Core)
4. Any further classification on the __System________________________
nature of the project : (System / Product / Component / Process)
5. Constraints considered : ______________________________________
(Details of the standards referred like: Economic, Environmental, Sustainability,
Manufacturability, Ethical, Health and Safety, Social, Political etc…. How this category
influenced the project is to be mentioned. Relevant page numbers may be marked)

Constraint /
Standard Remarks

6. Major Courses Covered in the Deep Learning, Machine Learning. Neural

Project : Networks.
7. SDGs Mapped :

SDG Name SDG No Supporting Remarks

Signature of the Student Signature of the Supervisor

v
 ABSTRACT

Keywords:

vi
 TABLE OF CONTENTS

CHAPTER PAGE
TITLE
NO. NO.
CERTIFICATE ii
DECLARATION iii
ACKNOWLEDGEMENT v
ABSTRACT vi
LIST OF TABLES xiii
LIST OF FIGURES xv
LIST OF ABBREVIATIONS xvii
1 INTRODUCTION 1
1.1 1
1.2 2
1.2.1 2
1.2.2 2
2 LITERATURE SURVEY
3 PROBLEM DEFINITION AND METHODOLOGY
4 RESULTS AND DISCUSSION
5 CONCLUSION AND FUTURE WORK
8.1 Conclusions Derived from The Contributions
8.2 Future Works and Enhancements
BIBLIOGRAPHY 103

vii
 LIST OF TABLES

Table Page
TITLE
No. No.

viii
 LIST OF FIGURES

Page
Figure No. Title
No.

ix
LIST OF ABBREVIATIONS

x
 CHAPTER 1
INTRODUCTION
1.1. GENERAL

The theoretical foundation and underlying concepts are explained in detail in

this chapter. Network Intrusion Detection Systems [1], Deep Learning [2], Machine
Learning [3], Auto Encoders [4], and Long Short-Term Memory [5] are discussed.

The purpose of a Network Intrusion Detection System (NIDS) is to monitor a

network for malicious activity, such as hacking [7], data theft, and other forms of
infiltration [6]. System administrators are alerted to any unusual behaviour in the
network or other data [8].

Machine Learning (ML) is a subfield of AI in which computers are taught to

draw conclusions about a problem from available data without being explicitly
programmed to do so. The ability to establish accurate patterns and predictions is
useful in many fields, not only in intrusion detection [10].

Deep Learning (DL) uses neural networks to model intricate connections

between inputs and outputs [11]. Image and audio recognition [12], natural language
processing [13], and intrusion detection are only some jobs that benefit significantly
from DL algorithms because of their ability to learn from vast volumes of data.

Auto Encoders (AE) are a sort of neural network to learn a compact

representation of the input data via unsupervised learning. They are used in this study
for data feature extraction.

The Recurrent Neural Network (RNN) [14] utilized for processing [15] is
called Long Short-Term Memory (LSTM). In this study, we employ LSTM, a
powerful method for modeling time-series data like network traffic, to categorize the
data.

11
1.2. WHAT IS NIDS BACKGROUND OF THE STUDY
1.2.1. Definition

A NIDS, or Network Intrusion Detection System, is a piece of security

software that can detect and report possible security breaches in a network. Its
purpose is to safeguard systems against being compromised by spies or malicious
users. Dorothy E. Denning [16] is widely regarded as an early innovator in the NIDS
area owing to her work at Purdue University in the late 1980s. Her 1986 article, "An
Intrusion Detection Model," published in the Journal of Software Engineering and
Methodology from IEEE, is a landmark publication.

1.3. PROBLEM DEFINITION

1.4. Motivation

1.5. Problem Statement

1.6. Limitations of the Existing System

1.7. Challenges

1.8. Objectives

1.9. Organization of the Thesis

1.10. Summary

Figure 1.1 The classification of IDS detection techniques and sources of data [31]

12
 CHAPTER 2
LITERATURE SURVEY
2.1. EXISTING RESEARCH

This literature study aims to examine autoencoders, LSTM, and RF [41]

applications in intrusion detection systems. The ability to analyze massive and
intricate data sets has contributed to the rise in the use of these intrusion detection
methods in recent years. For instance, autoencoders are well-suited for spotting
anomalies in network traffic, whereas LSTM excels in analyzing time-series data. The
RF method, on the other hand, is an effective ensemble way to label network data.
This study aims to comprehensively review current research by critically examining
the appropriate literature and elaborating on its significant contributions, limitations,
and potential future directions.

2.2. KEY FINDINGS

The accuracy with which intrusion detection systems (IDS) can identify
threats to a network has been greatly increased by using deep learning technology.
With their ability to process massive amounts of data, deep networks have emerged as
A literature review of unsupervised pre-trained networks is shown in Table 2.1.

13
 Table 2.1 Literature survey on unsupervised pre-trained networks

Model Author Year Description Advantages

A generative model
Can generate
Variational that uses a variational
Chen, Kingma, new samples
Autoencoder 2016 inference approach for
et al., [43] similar to the
(VAE) learning the encoder-
input data
decoder functions
An autoencoder that
Improves
Sparse uses sparsity
interpretability
Autoencoder Liu et al., [44] 2014 constraints on the
and reduces
(SAE) hidden layer
overfitting
activations
An autoencoder that Improves
Contractive enforces contractive robustness
Rifai, et al.,
Autoencoder 2011 constraints on the against minor
[45]
(CAE) hidden layer changes in the
activations input data
Ideal for image
An autoencoder
and audio data,
Convolution architecture that uses
Krizhevsky, et as it can
Autoencoder 2009 convolutional layers
al., [46] preserve the
s (CAE) for encoding and
spatial
decoding
information
Can produce a
Use numerous deep
Stacked
autoencoders for representation
Denoising Vincent, et al.,
2008 denoising, with each of the data,
Autoencoder [47]
autoencoder's output leading to
s (SDA)
feeding into the next. improved
performance
Improves
A multi-layer
representation
Hinton and autoencoder that can
Deep learning,
Salakhutdino 2006 capture high-level
Autoencoder leading to
v [48] features of the input
better accuracy
data
for some tasks
Autoencoder has the Improves
Denoising added ability to de- robustness
Soulié, et al.,
Autoencoder 1987 noise incoming data against noise
[49]
(DAE) and reconstruct the and corruption
original signal. in data
It can be used
A neural network
for
approach trained to
Bourlard, et dimensionality
Autoencoder 1987 replicate its input to
al., [50] reduction and
output through encoder
data
and decoder functions
compression

14
2.3. SUMMARY

Write summary of literature reviews. Write summary of literature reviews.

Write summary of literature reviews. Write summary of literature reviews. Write
summary of literature reviews.

15
 CHAPTER 3

METHODOLOGY

3.1. OVERVIEW

NIDS is developed to handle the challenge of detecting and preventing

network threats. NIDS is a security system that analyses network traffic for unusual
activities, such as malicious or unauthorized access. These attacks can take many
forms, such as unauthorized access attempts, denial of service attacks, malware
infections, etc. NIDS aims to identify these attacks as early as possible and alert
security personnel or take other automated actions to stop and protect the network.

3.2. MOTIVATION
3.3. LIMITATIONS OF THE EXISTING SYSTEM
 The huge growth in the volume of network data requires efficient techniques
to analyze data increasingly rapidly, efficiently, and effectively so that
sensitive data will be handled appropriately.
3.4. OBJECTIVES
 To convert the large volume of network data with mixed values (string and
numeric) into numeric using OneHotEncoder
 To provide an improved intrusion detection model without any information
loss using the LightGBM feature selection algorithm so that the efficiency and
accuracy of the system are improved

3.5. METHODOLOGY
3.5.1. Data Collection

3.5.2. Data Preprocessing

3.5.3. System Architecture

16
 Figure 3.1 Workflow of the proposed IDS model

3.5.4. Implementation Details

3.5.5. Algorithm

3.6. DATASETS USED

3.7. PERFORMANCE METRICS

A Network Intrusion Detection System's effectiveness may be measured in a

number of different areas. Some instances are as follows:

Accuracy: Network intrusion detection system (NIDS) performance is often

measured in part by how well it can identify actual intrusions. It is determined using
the following formula and assesses the proportion of NIDS forecasts that turn out to
be correct.

(True Positives+True Negatives)

Accuracy=
(True Positives+True Negatives+ False Positives+ False Negatives)
(3.1)

17
 Precision: This statistic may be calculated using the following formula, and it
indicates the percentage of positive cases recognized by the NIDS out of the total
number of positive predictions.

(True Positives)
Precision= (3.2)
(True Positives+ False Positives)

Recall: The NIDS's predictive accuracy is expressed as a percentage of the

total number of positive instances for which a prognosis was accurate. With the
following formula, this measurement may be determined.

(True Positives)
Recall= (3.3)
(True Positives+ False Negatives)

F1-score: This metric shows the harmonic mean of accuracy and recall and is
an excellent compromise. This is the calculation formula:

2∗(Precision∗Recall)
F 1−score= (3.4)
(Precision+ Recall)

False Positive Rate (FPR or fall-out): With this measure, we can determine
how often the NIDS incorrectly predicts a good outcome compared to how often it
correctly predicts a negative outcome. This is the formula that is used to determine
this metric:

False Positivession
FPR= (3.5)
(False Positives+ True Negatives)

If the NIDS has a low FPR, it is not producing many false alarms; if it has a high
FPR, it makes many false alerts. A high FPR in NIDS may be troublesome since it
leads to many false positives, which can cause security staff to get overwhelmed and
make it difficult to distinguish between false positives and actual intrusions.

3.8. SUMMARY

To better identify known and undiscovered assaults in large-scale, highly

dynamic, and complex network settings, researchers and developers must investigate,
build, and create novel methods; this is the challenge description for Network
Intrusion Detection Systems (NIDS).

18
 CHAPTER 4

RESULTS AND DISCUSSION

4.1. OVERVIEW

It is frequently necessary to undertake data preprocessing to improve the

quality and applicability of a dataset before analyzing it using machine learning
techniques. This preprocessing stage typically includes feature engineering, data
cleaning, and scaling tasks. One popular preprocessing technique is OneHotEncoding,
which transforms categorical data into numerical form. Additionally, normalization,
which standardizes numerical data, is another frequent technique. These preprocessing
methods ensure that the data is in a proper format for analysis, which could lead to
improved machine learning model performance.

4.2. EXPERIMENTAL SETUP

For setting up the experimental environment of NIDS using deep learning

techniques, Python programming

4.3. COMPARATIVE ANALYSIS

4.4. SUMMARY

19
 CHAPTER 5

CONCLUSION AND FUTURE WORK

8.1. CONCLUSIONS

8.2. FUTURE WORK

20
 BIBLIOGRAPHY

[1] M. Ring, S. Wunderlich, D. Scheuring, D. Landes, and A. Hotho, “A survey of

network-based intrusion detection data sets,” Computers and Security. 2019.
doi: 10.1016/j.cose.2019.06.005.
[2] A. Thakkar and R. Lohiya, “A Review of the Advancement in Intrusion
Detection Datasets,” Procedia Comput. Sci., vol. 167, pp. 636–645, Jan. 2020,
doi: 10.1016/J.PROCS.2020.03.330.
[3] R. C. Aygün and A. G. Yavuz, “A ˘ g Anomali Tespiti için Stokastik Veri Ayrı ¸
stırma Tabanlı Otomatik Kodlayıcı Yakla ¸ sımı A Stochastic Data
Discrimination Based Autoencoder Approach For Network Anomaly
Detection,” pp. 6–9, 2017.
[4] N. Shone, T. N. Ngoc, V. D. Phai, and Q. Shi, “A Deep Learning Approach to
Network Intrusion Detection,” IEEE Trans. Emerg. Top. Comput. Intell., 2018,
doi: 10.1109/TETCI.2017.2772792.

21