4.7 Article

P2CA-GAM-ID: Coupling of probabilistic principal components analysis with generalised additive model to predict the k-barriers for intrusion detection

Journal

Publisher

PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.engappai.2023.107137

Keywords

Machine learning; Intrusion detection; WSNs; Probabilistic PCA; GAM

Ask authors/readers for more resources

This research introduces a hybrid framework that combines P2CA and GAM algorithms, performing well in various WSN scenarios. The framework accurately predicts the number of barriers in three intrusion detection datasets and outperforms benchmark algorithms. It also demonstrates good adaptability and robustness by accurately predicting response variables in unrelated datasets. The findings have significant implications for reliable network security and protection of sensitive data and critical infrastructure.
Drastic advancement in computing technology and the dramatic increase in the usage of explainable machine learning algorithms provide a promising platform for developing robust intrusion detection algorithms. However, the development of these algorithms is constrained by their applicability over specific scenarios of Wireless Sensor Networks (WSNs). We introduced a hybrid framework by combining Probabilistic Principal Component Analysis (P2CA) and Generalised Additive Model (GAM), which is performing well for all the scenarios of WSNs. To demonstrate our framework's broad applicability, we evaluated its performance over three publicly available intrusion detection datasets (i.e., LT-FS-ID, AutoML-ID, and FF-ANN-ID), each from different scenarios. Our findings highlight that the presented framework can accurately predict the number of k-barriers for all three datasets. Furthermore, we conducted a comprehensive performance comparison between our proposed framework and benchmark algorithms, which revealed that our approach outperforms all of them. Additionally, we evaluated the framework's versatility by testing its performance on datasets unrelated to intrusion detection, specifically ALE datasets. Notably, our approach accurately predicted the response variable in these datasets and exceeded the performance of its primary algorithm, further demonstrating its robustness and adaptability. The implications of this research are substantial. By developing a robust intrusion detection framework that performs well across diverse WSN scenarios, we address a critical need for reliable network security in various domains, including industrial IoT, smart cities, and environmental monitoring. Our findings not only enhance the understanding of intrusion detection in WSNs but also pave the way for developing more sophisticated and adaptable systems to safeguard sensitive data and critical infrastructure.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.7
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available