Search

The emerging Internet of Things (IoT) is rapidly changing people’s life and work by integrating massive data collected by various smart devices. Thus, privacy protection for heavy IoT-device users becomes one of the most concerned problems in this field. However, it is found that a simple IoT device may give away people’s privacy by IoT-based data mining methods in this paper. It shows that with the help of transfer learning based on the energy disaggregation data from the IoT-based smart grid, even ordinary low-frequency grid load data could effectively reveal the resident’s privacy like family income or family member’s age. The results show that the pre-trained feature extractors for people’s electricity consumption patterns can help infer their privacies, which in turn proves that people’s private traits will affect their usage patterns of the electrical appliances. The results also warn that privacy protection should be concerned even by people who use only the simplest and the most common IoT devices (for example, a smart electric meter, which is used by more than 55% of families in the U.S.). [ABSTRACT FROM AUTHOR]