Proposal Submission – 4th April 2025

Today, I finalised and submitted my official Final Year Project (FYP) proposal titled “Comparative Analysis of Unsupervised Machine Learning Models for Network Intrusion Detection Systems” to the PSM1 evaluators.

Summary of Proposal Content:

• Research Focus:

  • Enhance Network Intrusion Detection Systems (NIDS) to detect unknown threats, including zero-day attacks.

  • Integrate and evaluate three unsupervised machine learning algorithms: Isolation Forest (IF), Local Outlier Factor (LOF), and K-Means Clustering.

• Problem Background:

  • Traditional NIDS relies on known signatures and is ineffective against novel attacks.

  • Manual updates and limited adaptability make it less scalable and efficient.

  • Unsupervised ML algorithms offer improved detection of unknown anomalies without requiring labeled data.

• Proposed Solution:

  • Simulate NIDS using unsupervised ML algorithms.

  • Compare the models’ ability to detect malicious packets in network traffic.

  • Evaluate and determine which algorithm performs best using standardized metrics.

• Research Objectives:

  1. Study the limitations of traditional NIDS and propose ML-based alternatives.

  2. Select suitable unsupervised ML algorithms and datasets.

  3. Design and simulate lightweight NIDS using the selected models.

  4. Evaluate performance using:

     • Precision

     • Recall

     • F1-Score

     • AUC-ROC

• Scope:

  • Algorithms: Isolation Forest, Local Outlier Factor, K-Means Clustering.

  • Datasets: CSE-CIC-IDS2018 and UNSW-NB15.

  • Tools: Python, Google Colab, Scikit-learn, NumPy, Pandas, Matplotlib, Seaborn.

  • Offline simulation only; no deep learning or real-time traffic detection.

• Project Area:

• Artificial Intelligence: Unsupervised ML

• Security: Network Intrusion Detection System (NIDS)

• Network: Network traffic simulation

Updated FYP Mindmap:

Achievements:

• Completed full proposal following PSM1 formatting and content requirements.

• Successfully submitted proposal before the deadline with supervisor approval.

• Defined a clear, feasible, and relevant research direction.

Upcoming Plans:

• Wait for proposal review results and respond to any required corrections.

• Begin outlining Chapters 1 and 2 for Progress Report 1.

• Expand literature review on chosen algorithms and datasets.