Lightweight Anomaly Detection Scheme Using Incremental Principal Component Analysis and Support Vector Machine.

Area of Science:

• Computer Science
• Network Engineering
• Data Science

Background:

• Wireless Sensor Networks (WSNs) are crucial for data collection across diverse fields like smart cities and healthcare.
• Ensuring data accuracy and reliability in WSNs is vital for decision-making, yet raw data is often imperfect.
• Detecting anomalies in resource-constrained WSNs presents a significant challenge due to computational and memory limitations.

Purpose of the Study:

• To design and develop a lightweight anomaly detection scheme for WSNs.
• To improve network efficiency by reducing computational complexity, communication overhead, and memory utilization.
• To maintain high accuracy in anomaly detection despite resource constraints.

Main Methods:

• Employed one-class learning and dimension reduction techniques.
• Utilized the One-Class Support Vector Machine (OCSVM) with a Centred-Ellipsoid kernel for anomaly detection.
• Implemented Candid Covariance-Free Incremental Principal Component Analysis (CCIPCA) for data dimension reduction.

Main Results:

• The proposed scheme achieved an anomaly detection accuracy exceeding 98%.
• Demonstrated efficient memory utilization with O(nd) complexity.
• Showcased no communication overhead, significantly outperforming existing schemes.
• Reduced computational complexity and memory footprint.

Conclusions:

• The developed lightweight anomaly detection scheme is effective and efficient for WSNs.
• The integration of OCSVM and CCIPCA offers a robust solution for resource-constrained environments.
• The scheme successfully balances high detection accuracy with minimal resource utilization.