Feature-Selection and Mutual-Clustering Approaches to Improve DoS Detection and Maintain WSNs' Lifetime

Area of Science:

• Computer Science
• Network Security
• Machine Learning

Background:

• Wireless Sensor Networks (WSNs) face significant energy and security challenges, particularly from Denial of Service (DoS) and Distributed DoS (DDoS) attacks.
• Traditional security methods are insufficient for WSNs due to increasing data complexity and the trend towards open-field encryption.
• Machine learning offers a viable solution for detecting sophisticated attacks in WSNs.

Purpose of the Study:

• To improve DoS anomaly detection and power reservation in WSNs.
• To evaluate the effectiveness of feature selection techniques combined with machine learning for WSN security and longevity.
• To introduce and assess the CH_Rotations clustering algorithm for enhanced anomaly detection efficiency.

Main Methods:

• A novel clustering algorithm, CH_Rotations, was developed to improve anomaly detection over the WSN's lifespan.
• Feature selection techniques, including Water Cycle (WC), Particle Swarm Optimization (PSO), Simulated Annealing (SA), Harmony Search (HS), and Genetic Algorithm (GA), were evaluated.
• Machine learning classifiers, specifically Decision Tree (DT), were used in conjunction with feature selection to analyze WSN node traffic.

Main Results:

• The Water Cycle (WC) feature selection method demonstrated superior performance, achieving higher accuracy than PSO, SA, HS, and GA.
• WC combined with the Decision Tree (DT) classifier achieved 100% accuracy in DoS detection using only a single feature.
• The CH_Rotations algorithm extended WSN network lifetime by 30% compared to the standard LEACH protocol.

Conclusions:

• Feature selection techniques, particularly WC, are crucial for effective DoS detection in WSNs, enabling high accuracy with minimal features.
• The CH_Rotations algorithm significantly enhances WSN network lifetime and anomaly detection efficiency.
• Combining advanced feature selection (WC) with machine learning (DT) offers a powerful strategy for securing WSNs while optimizing energy consumption.