A novel machine learning model for perimeter intrusion detection using intrusion image dataset
Perimeter Intrusion Detection Systems (PIDS) are crucial for protecting any physical locations by detecting and responding to intrusions around its perimeter. Despite the availability of several PIDS, challenges remain in detection accuracy and precise activi…
## Perimeter Intrusion Detection Using Machine Learning### Data Preprocessing- Duplicate images are removed for dataset integrity.- Data augmentation techniques like geometric transformation, kernel filters, and image mixing are applied to enhance the versatility of the deep learning-based image classification model.### Feature Extraction- Features are extracted from InceptionV3's penultimate layer using t-SNE for dimensionality reduction.- The idea behind Convolutional Neural Networks (CNNs) is explained, highlighting features like convolution operations and activation functions.### Enhanced DBSCAN- The study introduces an enhanced DBSCAN algorithm, automating parameter optimization through a novel approach to estimate the epsilon parameter.- Enhanced DBSCAN effectively identifies real intrusions and distinguishes them from non-intrusive scenarios, improving detection system accuracy.### The Proposed Model: Enhanced DBSCAN and Classification- The enhanced DBSCAN algorithm is applied to clustered data, and labeled images are used to classify various intrusion activities using k-NN classification.### Experimental Results- Comparative analysis with existing techniques shows improved performance for the proposed technique at higher ARI, FMI, and silhouette scores.- The model successfully distinguishes between intrusion types with high precision, recall, F1 score, and accuracy.### Summary and Future Work- The novel PIDS model leverages advanced data processing and machine learning techniques for enhanced perimeter intrusion detection.- It addresses challenges in PID by optimizing clustering and classification, resulting in a robust intrusion detection system.- Future research may incorporate a wider range of intrusion activity types and address intrusion detection from the top of perimeter fences.### Formatting Notes- HTML tags enclose the main content using `