Nowadays, it is more and more important to deal with the potential security issues of internet-of-things (IoT). Indeed, using the physical layer features of IoT wireless signals to achieve individual identity authentication is an effective way to enhance the security of IoT. However, traditional classifiers need to know all the categories in advance to get the recognition models. Realistically, it is difficult to collect all types of samples, which will result in some mistakes that the unknown target class may be decided as a known one. Consequently, this paper constructs an improving open-categorical classification model based on the generative adversarial networks (OCC-GAN) to solve the above problems. Here, we have modified the loss function of the generative model G and the discriminative model D. Compared to the traditional GAN model which can generate the fake sample overlapping with the real samples, our proposed G model generates the fake samples as negative samples which are evenly surrounding with the real samples, while the D model learns to distinguish between real samples and fake samples. Besides, we add auxiliary training not only to gain a better recognition result but also to improve the efficiency of the model. Furthermore, Our proposed model is verified through experimental study. Compared to other common methods, such as one-class support vector machine (OC-SVM) and one-versus-rest support vector machine (OvR-SVM), the OCC-GAN model has a better performance. The recognition rate of the OCC-GAN model can reach more than 90% with a recall rate of 97% by the data of the IoT module.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited