An efficient data driven framework for intrusion detection in wireless sensor networks using deep learning.

Area of Science:

• Cybersecurity
• Deep Learning
• Wireless Sensor Networks

Background:

• Wireless Sensor Networks (WSNs) are critical for distributed sensing but vulnerable to cybersecurity threats due to limited resources.
• Existing intrusion detection systems (IDS) struggle to address the unique challenges of WSNs.

Purpose of the Study:

• To design a robust and lightweight deep learning-based intrusion detection framework for WSNs.
• To enhance the detection accuracy and adversarial robustness of IDS in WSN environments.

Main Methods:

• A hybrid deep learning model combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs).
• An adversarial-aware optimization model with a compound objective function.
• Synthetic oversampling using SMOTE for data balancing.

Main Results:

• The proposed framework demonstrated superior performance across multiple benchmark datasets (NSL-KDD, CICIDS2017, UNSW-NB15, CTU-13).
• Achieved high detection accuracy while minimizing adversarial vulnerability and ensuring model generalizability.
• Exhibited strong robustness and transferability in cross-dataset and intra-dataset experiments.

Conclusions:

• The developed IDS is lightweight, resilient, and practical for WSN deployment.
• The deep learning approach offers a significant advancement in securing WSNs against sophisticated cyber threats.