Accurately and effectively identifying Wi-Fi devices is critical for improving network performance, ensuring security. In this paper, we aim to enhance the accuracy and efficiency of Wi-Fi device identification using over-the-air sniffing, which allows for device type identification without requiring knowledge of the network architecture. To achieve this, we design CFS-EDI, a novel system that integrates Comprehensive Feature Simplification (CFS) with an Ensemble Wi-Fi Device Identification model (EDI). Comprehensive Feature Simplification (CFS) analyzes the interaction traffic between devices and wireless access points, effectively enhancing Wi-Fi device identification accuracy and efficiency by eliminating redundant and noisy features and focusing on key attributes of device behavior. To effectively address the complexity and variability in Wi-Fi device traffic, we employ an Ensemble Wi-Fi Device Identification model (EDI) to enhance both the robustness and accuracy of Wi-Fi device identification. Ensemble Wi-Fi Device Identification model (EDI) consists of four base learners and a secondary classifier. Each base learner employs different learning strategies and algorithms, enhancing the system’s ability to capture diverse data patterns. The secondary classifier integrates the probability outputs of these base learners, leveraging their unique strengths and compensating for their individual weaknesses, significantly improving classification accuracy and robustness through comprehensive judgment. CFS-EDI is designed to analyze over-the-air Wi-Fi traffic without accessing the target network, minimizing resource consumption while maintaining high accuracy and efficiency. Extensive experiments on 30 different devices in real-world networks demonstrate that CFS-EDI achieves an accuracy of 98.13%, an F1 score of 98.14%, and a throughput of 355.35 packets per second, highlighting its high accuracy and efficiency.