With the proliferation of modern electromagnetic equipment and the rapid expansion of the Internet of Things (IoT), ensuring physical-layer security has become a critical concern. Specific Emitter Identification (SEI) methods are es-sential for identifying devices and protecting national security within the IoT environment. Despite the effectiveness of many radio frequency (RF) fingerprinting techniques based on deep learning (DL) and raw I/Q signals, current DL methods for SEI face limitations in learning short-term and long-term features simultaneously and fail to adaptively emphasize effective features. To address these challenges, this paper proposes a Multi-Sampling Convolutional Attention Network (MSCAN)-based SEI method. Our approach leverages a multi-sampling signal pro-cessing module to extract multi-scale features and incorporates a Squeeze-and-Excitation (SE) attention block to enhance the discriminative ability of these features, thereby improving the model's performance without increasing computational costs. The effectiveness of the proposed MSCAN method is validated on an open-source dataset, achieving an accuracy of 95.3%.