Unlike most of the upper layer authentication mechanisms, the physical (PHY) layer authentication takes advantages of channel impulse response from wireless propagation to identify transmitted packages with low-resource consumption, and machine learning methods are effective ways to improve its implementation. However, the training of the machine-learning-based PHY-layer authentication requires a large number of training samples, which makes the training process time consuming and computationally resource intensive. In this article, we propose a data augmented multiuser PHY-layer authentication scheme to enhance the security of mobile-edge computing system, an emergent architecture in the Internet of Things (IoT). Three data augmentation algorithms are proposed to speed up the establishment of the authentication model and improve the authentication success rate. By combining the deep neural network with data augmentation methods, the performance of the proposed multiuser PHY-layer authentication scheme is improved and the training speed is accelerated, even with fewer training samples. Extensive simulations are conducted under the real industry IoT environment and the figures illustrate the effectiveness of our approach.