Device Fingerprint Extraction Based on MFSJMI and NBC for Devices in Virtual Power Plant

With the rapid development of virtual power plant, which plays an important role in optimizing the layout of energy resources. However, the access authentication demand of mass volume terminal devices brings great security challenge. The access authentication relied on the device fingerprint attracts more and more attention. How to achieve stable and secure fingerprint extraction of enormous and various terminal devices in virtual power plant still need to be further study. Aiming to solve above problems, in this paper, we propose a device fingerprint extraction scheme based on the MFSJMI (Multi-label feature selection considering join mutual information and interaction weight) and the NBC (Naive Bayes Classifier). It takes the new feature relevance and the feature redundancy during the feature selection period. In addition, our scheme supporting the updating of device fingerprint library. Performance analysis indicates that the proposed scheme has more and better properties in aspects of stability, fingerprint library updating, fingerprint confidentiality. It is more feasible for being applied in virtual power plants.