Leveraging Deep Learning for IoT Transceiver Identification.

Area of Science:

• Cybersecurity
• Wireless Communication
• Machine Learning

Background:

• Internet of Things (IoT) networks face significant security challenges due to inadequate identification and authentication methods.
• Radio frequency fingerprinting (RFF) offers a hardware-based security solution, mitigating risks like key leakage and reducing computational load.
• Existing RFF techniques often exhibit location dependency, limiting their practical application.

Purpose of the Study:

• To propose a novel Radio Frequency Fingerprinting (RFF) scheme to enhance Internet of Things (IoT) security.
• To address and overcome the location-dependence limitations found in current RFF identification methods.
• To leverage spectrograms and carrier frequency offset (CFO) as unique device identifiers.

Main Methods:

• A new RFF scheme was developed, utilizing the spectrogram and carrier frequency offset (CFO) of radio signals.
• A convolutional neural network (CNN) was employed as the machine learning classifier for device identification.
• The proposed method was evaluated using real-world data, testing its robustness against location and time variations.

Main Results:

• The proposed RFF scheme achieved an accuracy of 80% in identifying devices.
• The system demonstrated effectiveness even when training and testing data were collected on different days and at different locations.
• This represents a 13% improvement in accuracy compared to existing state-of-the-art RFF approaches.

Conclusions:

• The novel RFF scheme effectively addresses the location-dependence problem in device identification for IoT networks.
• The use of spectrograms and CFO, coupled with CNN, provides a robust and accurate method for IoT security.
• This research offers a promising advancement for securing the growing landscape of Internet of Things applications.