Maximum Skewness Projection for Efficient Anomaly Detection in High-Dimensional Data

Anomaly detection is a critical task in various domains, including network security, industrial monitoring, and medical diagnostics, where identifying rare and abnormal events is essential for mitigating risks and ensuring system reliability. However, traditional anomaly detection methods often struggle with the “curse of dimensionality” in high-dimensional datasets, where outliers may only exhibit abnormality along a few specific directions. To address this challenge, we propose a novel approach called Maximum Skewness Projection (MSP). MSP utilizes the sensitivity of skewness to outliers to identify a low-dimensional projection subspace where anomalies are significantly separated from normal instances. We evaluate MSP on two types of datasets: hyperspectral image data and key performance indicators (KPIs) from wireless communication base stations. Experimental results demonstrate that MSP outperforms state-of-theart methods in terms of detection accuracy and computational efficiency.