An efficient privacy-preserving multilevel fusion-based feature engineering framework for UAV-enabled land cover...

Area of Science:

• Remote Sensing
• Cybersecurity
• Artificial Intelligence

Background:

• Unmanned Aerial Vehicles (UAVs) offer significant advantages over manned aircraft for applications like land cover classification.
• Networked UAV systems are susceptible to malicious attacks, necessitating robust intrusion detection systems (IDSs).
• Deep learning (DL) is crucial for addressing security challenges in UAV networks.

Purpose of the Study:

• To propose an effective Privacy-Preserving Intrusion Detection Model for UAV-Based Remote Sensing Applications in Land Cover Classification Using Multilevel Fusion Feature Engineering (IDUAVRS-LCCMFFE).
• To enhance the accuracy and security of land cover classification using UAV imagery in dynamic environments.

Main Methods:

• Image pre-processing using a joint bilateral filter (JBF) to reduce noise.
• Multilevel feature extraction via fusion of NASNetMobile, ResNet50, and VGG19 models.
• Land cover classification using an Elman recurrent neural network (ERNN) optimized with the Salp Swarm Algorithm (SSA).

Main Results:

• The IDUAVRS-LCCMFFE model achieved superior accuracy, reaching 99.66% on the ToN-IoT dataset.
• Validation on the EuroSat dataset also demonstrated high performance with 96.47% accuracy.
• The proposed technique effectively enhances image quality and classification accuracy for UAV-based remote sensing.

Conclusions:

• The IDUAVRS-LCCMFFE model provides a robust solution for intrusion detection in UAV networks for land cover classification.
• The multilevel fusion feature engineering approach significantly improves classification performance.
• This research contributes to securing UAV-based remote sensing applications.