Research on the Classification and Recognition Algorithm of Intrusion Events Leveraging Machine Learning Methodologies

Fiber optic perimeter security systems have emerged as an effective means for protecting critical infrastructure and sensitive areas, owing to their high sensitivity, resistance to electromagnetic interference, and capability for longdistance monitoring. However, challenges such as false alarms, mis...

Full description

Saved in:
Bibliographic Details
Published in:2025 International Conference on Communication Networks and Smart Systems Engineering (ICCNSE) pp. 224 - 229
Main Authors: Ma, Yushu, Chen, Xi, Liu, Chengxin, Liu, Wei
Format: Conference Proceeding
Language:English
Published: IEEE 01.08.2025
Subjects:
Accuracy
Classification algorithms
Convolutional neural networks
Convolutional neural networks (CNNs)
Data models
Deep learning
Fiber optic security systems
Intrusion detection
Intrusion detection algorithm
Optical fibers
Security
Training
Vibrations
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Fiber optic perimeter security systems have emerged as an effective means for protecting critical infrastructure and sensitive areas, owing to their high sensitivity, resistance to electromagnetic interference, and capability for longdistance monitoring. However, challenges such as false alarms, missed detections, and performance degradation in complex environments necessitate the development of intelligent intrusion detection algorithms. The Support Vector Machine (SVM) algorithm is a powerful classifier. It trains the model by finding the hyperplane that distinguishes between two classes and maximizing the margin. It is applicable to the processing of small and mediumsized datasets and nonlinear problems. However, it has high computational costs, is sensitive to parameters, and is difficult to interpret the model. As a result, its application is limited in the scenarios of optical fiber perimeter security intrusion recognition that involve processing large-scale data and have high requirements for real-time performance. This study proposes designing an intrusion detection algorithm for fiber optic security systems to enhance event recognition accuracy. Using deep learning techniques, particularly convolutional neural networks (CNNs), the paper classifies and recognizes fiber optic vibration signals. Key research activities include data preprocessing, network architecture design, and the training and evaluation of the deep learning model. Experimental results demonstrate the algorithm's effectiveness in identifying various intrusion events, it has effectively solved the problem of insufficient accuracy in identifying intrusion events caused by false alarms, missed alarms and performance fluctuations in complex environments in the practical application of fiber optic perimeter security systems, and improved the system's precise discrimination ability for different intrusion behaviors.
DOI:10.1109/ICCNSE66404.2025.11144179