In the Internet of Things (IoT) era where billions of connected devices interact within smart environments, ensuring secure and accurate identification of these devices is paramount. Device fingerprinting has emerged as a crucial technique to uniquely identify and monitor devices without relying on traditional authentication methods. Given the proliferation of resource-constrained IoT devices, non-intrusive and scalable security solutions are essential. Unlike conventional identification mechanisms, device fingerprinting leverages unique device characteristics to provide a robust security layer. Over the past few years, machine learning (ML) techniques have increasingly been integrated into device fingerprinting methods. These approaches can detect subtle variations in device behaviour and attributes, making them indispensable for device recognition in dynamic IoT environments. In this paper, we conduct a comparative analysis of six representative ML-based device fingerprinting methods using publicly available benchmark datasets. The performance of these methods is evaluated using several metrics, including identification accuracy, time overhead, CPU consumption, and memory usage. Our analysis shows that IoTPROFILE outperforms other methods in performance effectiveness, while IoTDevID and IoT SENTINEL excel in computational efficiency. This provides insights into the strengths and weaknesses of each method, offering a clear understanding of their applicability in different contexts and environments.