Signal Processing for Time Domain Wavelengths of Ultra-Weak FBGs Array in Perimeter Security Monitoring Based on...

Area of Science:

• Computer Science
• Artificial Intelligence
• Cybersecurity

Background:

• Perimeter intrusion detection systems (PIDS) require efficient real-time data processing.
• Traditional density-based spatial clustering of applications with noise (T-DBSCAN) relies on manual parameter adjustments, limiting its real-time application.
• Stream computing offers potential for enhanced data processing in PIDS.

Purpose of the Study:

• To develop an improved real-time data processing method for perimeter intrusion detection systems.
• To introduce an unsupervised clustering algorithm that eliminates manual parameter tuning.
• To implement and evaluate the proposed method on a stream computing platform.

Main Methods:

• Implementation of an adaptive parameters DBSCAN (AP-DBSCAN) algorithm for unsupervised calculations.
• Integration of the AP-DBSCAN method with the Spark Streaming platform for data stream collection and real-time analysis.
• Development of capabilities for judging and identifying different types of intrusions.

Main Results:

• The AP-DBSCAN method demonstrated effective adaptive parameter calibration on the Spark Streaming platform.
• The proposed method achieved accuracy comparable to T-DBSCAN without manual parameter setting.
• The system showed good overall performance in perimeter intrusion detection scenarios.

Conclusions:

• The AP-DBSCAN method on Spark Streaming provides an accurate and efficient solution for real-time perimeter intrusion detection.
• Unsupervised parameter adaptation enhances the usability and effectiveness of DBSCAN in dynamic environments.
• This approach addresses key challenges in data stream processing for security applications.