A Security Model for the Classification of Suspicious Data Using Machine Learning Techniques - CNN, LSTM, RNN, GRU, and MLP DNN

Aug 3, 2026

Cybercrime has emerged as a persistent global threat, amplified by widespread internet use and dependence on technology. Countries vary in attack types and defensive capacity, making tailored and resource-aware mitigation essential. Although sophisticated machine-learning-driven attacks are less common in some settings, cybercrime remains a costly worldwide challenge. This study builds a security approach for suspicious data classification using five machine learning models—CNN, LSTM, RNN, GRU, and MLP DNN—trained and tested on the InSDN public dataset, yielding different training and testing accuracy levels.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages6

Reading Time15 min

AccessFree to download

Abstract

Introduction

Cybercrime overview

Role of machine learning and AI in cybersecurity

Types of cyber-attacks

Challenges in attribution

Crime categorization

Machine learning background

Convolutional Neural Networks (CNN) basics

A Security Model for the Classification of Suspicious Data Using Machine Learning Techniques Boussi Grace Odette1, Himanshu Gupta2, Syed Akhter Hossain3 1Ph.D. Scholar, AIIT Amity University Noida, India [graceboussi@gmail.com](graceboussi@gmail.com) 2Professor, AIIT Amity University Noida, India [hgupta@amity.edu](hgupta@amity.edu) 3Professor, Computer Science and Engineering Department University of Liberal Arts Dhaka, Bangladesh [aktarhossain@daffodilvarsity.edu.bd](aktarhossain@daffodilvarsity.edu.bd) Abstract—Cybercrime first emerged in 1981 and gained significant attention in the 20th century. The proliferation of technology and our increasing reliance on the internet have been major factors contributing to the growth of cybercrime. Different countries face varying types and levels of cyber-attacks, with developing countries often dealing with different types of attacks compared to developed countries. The response to cybercrime is usually based on the resources and technological capabilities available in each country. For example, sophisticated attacks involving machine learning may not be common in countries with limited technological advancements. Despite the variations in technology and resources, cybercrime remains a costly issue worldwide, projected to reach around 8 trillion by 2023. Preventing and combating cybercrime has become crucial in our society. Machine learning techniques, such as convolutional neural networks (CNN), recurrent neural networks (RNN), and more, have gained popularity in the fight against cybercrime. Researchers and authors have made significant contributions in protecting and predicting cybercrime. Nowadays, many corporations implement cyber defense strategies based on machine learning to safeguard their data. In this study, we utilized five different machine learning algorithms, including CNN, LSTM, RNN, GRU, and MLP DNN, to address cybercrime. The models were trained and tested using the InSDN public dataset. Each model provided different levels of trained and test accuracy percentages. Keywords-cybercrime, cybersecurity algorithm, CNN, RNN, GRU, LSTM, machine learning, model. I. INTRODUCTION Cybercrime refers to illegal activities that are committed online using electronic devices. It has become a widespread and well-known issue, posing a major challenge globally. Cybercriminals target various sectors, with the financial industry being a prime focus due to the potential for monetary gain. To protect sensitive data and prevent cyber-attacks, cybersecurity experts employ various techniques and methods. One effective approach is the use of machine learning (ML) and artificial intelligence (AI). These technologies have proven to be valuable tools in handling the complexities of cyber-attacks. ML and AI algorithms and models have been integrated into existing security systems to enhance their effectiveness. These technologies have shown promising results in detecting and mitigating cyber threats. However, as financial organizations implement AI and ML for their own security, cybercriminals also leverage these tools to make their attacks more sophisticated and harder to detect [4] . Cyber-attacks result in significant financial losses for both countries and individuals on a daily basis. There is a wide range of cybercrimes orchestrated through the internet, including ransomware attacks and phishing scams [10] . Ransomware involves encrypting victims' data and demanding a ransom for its release. Phishing, on the other hand, is a prevalent form of social engineering where attackers trick individuals into revealing sensitive information [1] . Tracking and attributing the components ofa cyber-attack toa specific threat actor pose significant challenges for cybersecurity systems [6] . Cybersecurity experts continuously work to improve their techniques in order to detect, prevent, and respond to cyber threats effectively. It is important to note that

FAQs

What problem does the document address?

It addresses classification of suspicious data to improve cybersecurity against cybercrime and cyber-attacks.

Which machine learning algorithms are used in the proposed security model?

The study evaluates CNN, LSTM, RNN, GRU, and MLP DNN models for suspicious data classification.

What dataset is used to train and test the models?

Models are trained and tested using the InSDN public dataset.

Aug 3, 2026

Upload by Olivia Brown