Federated learning with LSTM for intrusion detection in IoT-based wireless sensor networks: a multi-dataset analysis

Area of Science:

• Cybersecurity
• Machine Learning
• Wireless Sensor Networks

Background:

• Internet of Things (IoT) and Wireless Sensor Networks (WSNs) are vulnerable to security breaches.
• Traditional centralized Intrusion Detection Systems (IDS) struggle with privacy, efficiency, and scalability in IoT environments.

Purpose of the Study:

• To develop and evaluate a federated learning (FL) framework integrated with long short-term memory (LSTM) networks for intrusion detection in IoT-based WSNs.
• To enhance detection accuracy, minimize false positive rates (FPR), and ensure data privacy while maintaining scalability.

Main Methods:

• A federated learning (FL) approach was used, enabling collaborative training of a global LSTM model across multiple IoT nodes without raw data exchange.
• The proposed FL-based LSTM model was tested on WSN-DS, CIC-IDS-2017, and UNSW-NB15 datasets.

Main Results:

• The FL-based LSTM model demonstrated significant improvements in intrusion detection compared to traditional centralized models.
• Higher accuracy and lower FPR were achieved across all tested datasets.
• The model effectively handled sequential data, ensuring privacy and maintaining performance in complex attack scenarios.

Conclusions:

• Federated learning and LSTM networks offer a robust solution for intrusion detection in IoT-based WSNs, enhancing both privacy and detection capabilities.
• The proposed framework addresses key IoT security challenges, including data privacy, scalability, and performance, making it suitable for real-world applications.