Abstract
Traditional device fingerprinting methods primarily rely on data-intensive deep learning models, treating devices as black boxes. However, these approaches suffer from high computational complexity and sensitivity to noise, making them impractical for large-scale Internet-of-Things (IoT) deployments that demand real-time processing and energy efficiency. To address these challenges, we propose a physics-based fingerprinting model for optical wireless communication (OWC) systems. Instead of relying solely on statistical features, small yet significant non-linear response variations in LEDs are quantified as lumped parameters in an equivalent circuit model, serving as the foundation for physical fingerprint features. This physics-based fingerprint is not only interpretable but also enables compact feature representation, reducing reliance on extensive training data. Experimental results demonstrate that our method achieves an average classification accuracy of 90.88% under varying signal-to-noise ratio (SNR) conditions. Furthermore, compared to convolutional neural network (CNN) and long short-term memory (LSTM)-based models, our approach achieves higher accuracy with significantly lower computational overhead, making it particularly well-suited for resource-constrained IoT environments. By reducing both dataset dimensionality and training sample requirements, our framework provides an efficient and scalable solution for device authentication in OWC networks.