A Comparative Analysis of Machine Learning and Hybrid Approaches for Network Intrusion Detection: Performance Evaluation and Implementation Framework
Fasanmi Ezekiel Olufemi, Department of Computer Science, Delta State University, Abraka, Nigeria
Published in Journal of Emerging Trends and Novel Research (JETNR), Volume 14, Issue 4, August 2025 | ISSN: 2984-9276 | JETNR53435
Network Intrusion Detection Systems (NIDS) represent a critical component of modern cybersecurity infrastructure, serving as the first line of defense against sophisticated cyber threats. This research presents a comprehensive analysis of current network intrusion detection methodologies, evaluating their effectiveness in detecting various attack patterns. We conducted a systematic review of signature-based, anomaly-based, and machine learning approaches to intrusion detection, analyzing their performance across multiple datasets including DARPA, NSL-KDD, and CICIDS2017. Our methodology involved implementing and testing various detection algorithms, measuring their accuracy, false positive rates, and computational efficiency. Results indicate that hybrid approaches combining signature-based detection with machine learning techniques achieve the highest detection rates (94.7% accuracy) while maintaining acceptable false positive rates (2.3%). Deep learning models, particularly Convolutional Neural Networks, demonstrated superior performance in detecting zero-day attacks but required significantly more computational resources. The study reveals that while traditional signature-based systems remain effective for known threats, the integration of artificial intelligence techniques is essential for addressing evolving attack vectors and encrypted traffic analysis. These findings contribute to the understanding of optimal NIDS configurations for modern network environments and provide guidance for security practitioners in selecting appropriate detection strategies.
Deep Learning for Intrusion Detection in IoT Networks: A Scalable Approach
Dr. Aisha Bello, Department of Cybersecurity, University of Lagos, Nigeria
Published in Journal of Emerging Trends and Novel Research (JETNR), Volume 14, Issue 3, June 2025 | ISSN: 2984-9276 | JETNR53412
The proliferation of Internet of Things (IoT) devices has introduced new challenges for network security, necessitating advanced intrusion detection systems (IDS). This study explores the application of deep learning techniques, specifically Long Short-Term Memory (LSTM) networks, for detecting intrusions in IoT networks. Using the IoT-23 dataset, we evaluated the performance of LSTM models against traditional machine learning approaches. Results show that LSTM models achieve a detection accuracy of 92.3% with a false positive rate of 3.1%, outperforming conventional methods in identifying complex attack patterns. The study proposes a scalable framework for deploying deep learning-based IDS in resource-constrained IoT environments, addressing computational efficiency through model optimization. These findings highlight the potential of deep learning for enhancing IoT security and provide a roadmap for future research in adaptive IDS.
Anomaly Detection in Blockchain Networks Using Machine Learning
Prof. Chukwuemeka Obi, Department of Computer Engineering, Ahmadu Bello University, Zaria, Nigeria
Published in Journal of Emerging Trends and Novel Research (JETNR), Volume 14, Issue 2, April 2025 | ISSN: 2984-9276 | JETNR53389
Blockchain networks, while inherently secure, are increasingly targeted by sophisticated cyberattacks. This research investigates anomaly detection techniques for identifying malicious activities in blockchain networks using machine learning. We evaluated Isolation Forest and Autoencoder models on a custom blockchain dataset, achieving detection rates of 89.5% and 91.2%, respectively. The study highlights the effectiveness of unsupervised learning for detecting unknown threats in decentralized systems. Challenges such as high computational costs and data imbalance were addressed through feature engineering and synthetic data generation. The proposed framework offers practical insights for securing blockchain-based applications, with implications for financial and supply chain systems.
Encrypted Traffic Analysis for Network Intrusion Detection Using Machine Learning
Dr. Ngozi Adebayo, Department of Information Technology, Federal University of Technology, Akure, Nigeria
Published in Journal of Emerging Trends and Novel Research (JETNR), Volume 14, Issue 1, February 2025 | ISSN: 2984-9276 | JETNR53365
The increasing use of encrypted traffic poses significant challenges for network intrusion detection systems (NIDS). This study proposes a machine learning-based approach for analyzing encrypted traffic without decryption, focusing on metadata and traffic patterns. Using the CICIDS2017 dataset, we implemented Random Forest and Gradient Boosting models, achieving a detection accuracy of 93.8% for malicious encrypted traffic. The research emphasizes feature extraction techniques, such as packet size distribution and flow duration, to enhance detection performance. Results suggest that machine learning models can effectively identify threats in encrypted environments, offering a scalable solution for modern networks. Future work will explore real-time analysis and integration with zero-trust architectures.