An Enhanced Intrusion Detection Model Based on Improved kNN in WSNs

Area of Science:

• Computer Science
• Network Security
• Artificial Intelligence

Background:

• Wireless Sensor Networks (WSNs) are vulnerable to various cyber threats, necessitating robust security measures.
• Traditional intrusion detection systems (IDS) may face challenges in resource-constrained WSN environments.
• Edge computing offers a promising paradigm for deploying intelligent security solutions closer to data sources.

Purpose of the Study:

• To develop an intelligent intrusion detection model for WSNs leveraging edge computing.
• To enhance the detection accuracy and efficiency of WSN intrusion detection systems.
• To specifically address Denial-of-Service (DoS) attacks in WSNs.

Main Methods:

• Proposed an edge intelligence framework integrating the k-Nearest Neighbor (kNN) algorithm and the Arithmetic Optimization Algorithm (AOA).
• Introduced a Parallel strategy and Lévy flight strategy to optimize the AOA (PL-AOA) for enhanced kNN classifier performance.
• Utilized the WSN-DS dataset for simulation experiments in Matlab2018b.

Main Results:

• The proposed PL-AOA algorithm demonstrated strong performance in benchmark function tests.
• The intelligent intrusion detection model achieved 99% accuracy (ACC) in detecting DoS attacks.
• Achieved a nearly 10% improvement in accuracy compared to the standard kNN classifier for DoS intrusion detection.

Conclusions:

• The developed edge-based intelligent intrusion detection model is effective for WSN security.
• The integration of PL-AOA with kNN offers significant improvements in detecting DoS attacks.
• The proposed framework has practical application significance for securing WSNs.