A physics-based fingerprinting approach for efficient device identification in OWC system.

Area of Science:

• Electrical Engineering
• Computer Science
• Communication Systems

Background:

• Traditional device fingerprinting relies on complex deep learning models, unsuitable for resource-constrained Internet-of-Things (IoT) environments.
• Existing methods face challenges with high computational complexity and noise sensitivity, hindering real-time processing and energy efficiency.

Purpose of the Study:

• To develop an efficient and interpretable physics-based fingerprinting model for optical wireless communication (OWC) systems.
• To address the limitations of traditional black-box approaches in IoT device authentication.

Main Methods:

• Quantified non-linear LED response variations as lumped parameters in an equivalent circuit model.
• Developed a physics-based fingerprinting approach utilizing these parameters for feature extraction.
• Evaluated the model's performance under varying signal-to-noise ratio (SNR) conditions.

Main Results:

• Achieved an average classification accuracy of 90.88% under different SNR conditions.
• Demonstrated higher accuracy and significantly lower computational overhead compared to CNN and LSTM models.
• Showcased reduced dataset dimensionality and training sample requirements.

Conclusions:

• The proposed physics-based fingerprinting model is efficient and scalable for device authentication in OWC networks.
• This approach is well-suited for resource-constrained IoT environments demanding real-time processing and energy efficiency.
• Offers an interpretable and compact feature representation, reducing reliance on extensive training data.