The rapid proliferation of the Internet of Things (IoT) devices has significantly increased the complexity of IoT device identification and management. Currently, intelligent applications controlling IoT devices employ rapid configuration mechanisms to swiftly configure and authorize devices based on their types. However, they are vulnerable to device ID spoofing, such as MAC address spoofing. Illegitimate devices may impersonate legitimate ones to gain permissions, thereby posing significant security risks. Therefore, we propose a device identification system called DevDet to achieve precise device identification. In this study, we introduce a feature extraction method based on autoencoders and utilize DT algorithm to identify device features. DevDet achieves identification accuracy exceeding 0.98 on three datasets, significantly outperforming other comparative algorithms. Considering the potential exploitation of DevDet by attackers to infer the usage of household devices, we propose a traffic obfuscation scheme to mislead attackers and reduce the accuracy of device inference.