Effective zero-shot learning method for event classification in Φ-OTDR sensing systems.

Area of Science:

• Fiber Optic Sensing
• Machine Learning
• Intrusion Detection Systems

Background:

• Existing Φ-OTDR methods require extensive training data, leading to overfitting on common events.
• Difficulty in simulating or acquiring samples for rare intrusion events poses a significant challenge.
• This limits the robustness of fiber optic sensing intrusion detection systems.

Purpose of the Study:

• To propose a zero-sample learning model for recognizing one-dimensional intrusion events with insufficient training data.
• To address the overfitting issue common in traditional intrusion detection models.
• To enhance the performance of Φ-OTDR systems in detecting rare or novel intrusion events.

Main Methods:

• Development of a zero-sample learning one-dimensional residual model (APL-ZSL-1DResNet).
• Incorporation of attribute point loss (APL) to improve recognition accuracy with limited samples.
• Validation on both a self-made and an open dataset, treating each category as zero-sample events.

Main Results:

• Achieved an average recall rate of 75% and 66% for zero-sample intrusion events.
• Maintained high average recall rates of 94.6% and 83.5% for common intrusion events.
• Demonstrated the model's effectiveness in scenarios with scarce or difficult-to-simulate intrusion data.

Conclusions:

• The proposed APL-ZSL-1DResNet model significantly improves intrusion event recognition with insufficient training data.
• Zero-sample learning offers a viable solution to the limitations of data-hungry models in Φ-OTDR systems.
• This approach enhances the adaptability and reliability of fiber optic sensing for security applications.