Rogue device discrimination in ZigBee networks using wavelet transform and autoencoders

Area of Science:

• Computer Science
• Electrical Engineering
• Cybersecurity

Background:

• Modern wireless systems like ZigBee transmit sensitive data, increasing vulnerability to unauthorized access and malicious attacks.
• Rogue devices pose a significant threat by attempting to tamper with authorized device identities and compromise network security.
• Existing security measures require enhancement to effectively detect and reject spoofing or rogue devices in wireless networks.

Purpose of the Study:

• To develop a robust method for detecting and rejecting spoofing/rogue devices in ZigBee networks.
• To leverage Radio Frequency Distinct Native Attributes (RF-DNA) for unique device identification.
• To improve classification accuracy for distinguishing between legitimate and malicious devices.

Main Methods:

• Extraction of physical layer features, specifically preambles, from ZigBee devices to create a distinguishable dataset.
• Application of deep learning algorithms, including wavelet decomposition and autoencoder representation, for feature analysis.
• Development of classification models to identify and reject unauthorized or spoofing devices based on RF-DNA.

Main Results:

• Successful creation of a dataset based on RF-DNA, enabling the distinction of individual ZigBee devices.
• Achieved high classification accuracy in identifying spoofing/rogue devices through deep learning models.
• Demonstrated the effectiveness of wavelet decomposition and autoencoder representation in feature extraction and classification.

Conclusions:

• RF-DNA analysis combined with deep learning offers a promising approach for enhancing security in ZigBee networks.
• The proposed methods effectively address the challenges of feature extraction and efficient model design for device discrimination.
• This technique significantly improves the ability to detect and reject rogue devices, bolstering overall network integrity.