A Comparison of Feature Selection and Feature Extraction in Network Intrusion Detection Systems
Date:
I presented our research on comparing feature selection and feature extraction techniques in network intrusion detection systems at the 2022 APSIPA ASC conference in Bangkok, Thailand. This talk showcased our comprehensive evaluation of dimensionality reduction approaches for enhancing cybersecurity systems.
Presentation Overview
The presentation focused on our comparative analysis of different dimensionality reduction techniques in the context of network intrusion detection systems (NIDS). With the increasing complexity and volume of network traffic, effective feature processing has become essential for building efficient and accurate NIDS.
Key Topics Covered
- Introduction to Network Security Challenges:
- The growing complexity of network attacks
- Computational constraints in real-time network monitoring
- High-dimensional nature of network traffic data
- Methodology:
- Feature selection methods (Information Gain, Chi-Square, Correlation-based)
- Feature extraction approaches (PCA, LDA, Autoencoders)
- Experimental design and evaluation metrics
- Results and Analysis:
- Performance comparison across multiple datasets
- Trade-offs between computational efficiency and detection accuracy
- Impact on model interpretability and explainability
- Practical Implications:
- Guidelines for selecting appropriate techniques based on specific security requirements
- Implementation considerations for operational environments
- Future research directions in adaptive feature processing
Conference Experience
The presentation generated significant interest among cybersecurity researchers and practitioners at the conference. The Q&A session sparked valuable discussions around the practical applications of these techniques in real-world security operations centers. Several attendees expressed interest in potential collaborations and extensions of this work.
Being part of APSIPA ASC 2022 provided an excellent opportunity to engage with the international research community and gain insights into cutting-edge developments in signal processing and information security. The feedback received helped refine our approaches and inform subsequent research directions in this domain.