Generative adversarial local density-based unsupervised anomaly detection.

Area of Science:

• Computer Science
• Data Science
• Artificial Intelligence

Background:

• Anomaly detection is critical for financial fraud, cybersecurity, and health monitoring.
• Existing Generative Adversarial Nets (GANs) methods struggle with complex data distributions due to overlooking local density.

Purpose of the Study:

• To develop an advanced anomaly detection method that incorporates local density information.
• To improve the accuracy and robustness of anomaly detection in diverse, complex datasets.

Main Methods:

• Introduced the Generative Adversarial Local Density-based anomaly detection (GALD) method.
• Utilized GANs to model normal data distributions and generate synthetic data.
• Calculated local synthetic density to identify anomalies based on deviations from normal data neighborhoods.

Main Results:

• The GALD method achieved an average AUC of 0.874 and an accuracy of 94.34% across seven real-world datasets.
• Significantly outperformed seven state-of-the-art anomaly detection methods.
• Demonstrated superior performance in medical diagnostics, industrial monitoring, and material analysis.

Conclusions:

• GALD effectively combines GANs' data modeling with local density analysis for superior anomaly detection.
• The method shows strong potential for applications requiring high accuracy in complex data environments.