Volume 15, Issue 4, November 2024

Comparative Analysis of Machine Learning Classifiers for Resource-Constrained IoT Intrusion Detection

Muneeb Javed, Rizwan Ishaq

Department of Data Science, University of Engineering and Technology, Lahore-Pakistan

Abstract - Although the Internet of Things (IoT) has grown to be an essential component of contemporary infrastructure, it is nevertheless vulnerable to a wide range of security risks, the most disruptive of which are Denial of Service (DoS) assaults. Using a Python-based analytical stack that consists of Pandas, Seaborn, Matplotlib, Scikit-learn, XGBoost, LightGBM, AdaBoost, Gradient Boosting, Bagging Classifiers, and Keras, this study uses the BoTNeTIoT-L01 dataset to perform a comparative analysis of different machine learning (ML) classifiers. In order to provide a benchmark for IDS appropriate for the limited resources of IoT devices, we evaluate the classifiers' performance in terms of accuracy, precision, recall, F1-score, and computational efficiency. By weighing the trade-off between detection capabilities and system resource constraints, the study aims to provide guidance for the implementation of effective, real-time intrusion detection systems (IDS) in IoT networks.

Index Terms - Network Security, Intrusion Detection Systems (IDS), Internet of Things, Anomaly, Significance Test, Denial of Service, And TCP/IP Data and Performance Analysis

Enhancing IoT Devices Security by Integrating Zero Trust Model Using Machine Learning Techniques in a Public Cloud

Pages: 9-17 (9) | [Full Text] PDF (597K)

F. Zubair, Talha Majeed

Department of Computer Science, University of Engineering and Technology Lahore, Pakistan

Abstract - The rapid growth of the Internet of Things (IoT) has introduced significant security vulnerabilities, necessitating advanced methods for protecting IoT ecosystems. This thesis presents a novel approach to enhancing IoT security by integrating the Zero Trust model with machine learning techniques. Specifically, a CNN-BiLSTM-based Intrusion Detection System (IDS) is proposed to detect and mitigate various cyber threats, including Distributed Denial of Service (DDoS), spoofing, and Man-in-the-Middle (MitM) attacks, within a Zero Trust framework. The CICIDS2017 dataset was employed for model training and evaluation, ensuring comprehensive coverage of IoT-related attack vectors. A comparative analysis was conducted using three machine learning models—CNN-BiLSTM, Support Vector Machines (SVM), and Linear Regression—evaluating their performance based on accuracy, loss, and the computed trust values of IoT devices. The CNN-BiLSTM model outperformed the other models, achieving 100% test accuracy and generating the highest trust scores for IoT devices. These trust values were dynamically computed as part of the Zero Trust policy enforcement, ensuring that only authenticated and verified devices could access the system. This research also introduces a feedback loop, where continuous monitoring of device behavior and machine learning predictions feed back into the system to adjust policies dynamically, further strengthening the security posture. The results demonstrate the effectiveness of combining Zero Trust principles with machine learning for robust, real-time IoT security, offering an innovative solution for mitigating advanced cyber threats.

Index Terms - ML, Zero Trust Model (ZTM), IoT, CNN and Cyber Threats

Pages: - () | [Full Text] PDF (308K)

Abstract -

Index Terms -

Pages: - () | [Full Text] PDF (314K)

Abstract -

Index Terms -

Pages: - () | [Full Text] PDF (314K)

Abstract -

Index Terms -