Peiyu Chen, Zhaohui Hu, Qianjun Tu
Differential privacy; TOP-K query; Time automatic wave neural net-work; Frequency prediction; K-modes clustering
To address the privacy risks of high-frequency and fine-grained elec- tricity data collected by smart meters, which may expose user be- havior patterns and lifestyle habits, this study proposes a smart me- ter data protection model that integrates K-modes clustering–based shuffled differential privacy with a time auto-wave neural network (TANN) for TOP-K queries. In the data perturbation stage, the model allocates privacy budgets efficiently through frequency pre- diction and a gradient random response mechanism. In the query stage, it employs temporal dependency modeling within the TANN structure to enhance the real-time capability and accuracy of TOP-K queries. Experimental results show that the proposed model achieves a normalized intra-cluster variance of 0.23, an F1 score of 0.976, and a root mean square error of 0.155, indicating superior clustering per- formance. The TOP-K query time is only 4.6 seconds, with a mean absolute error of 4.5% and a re-identification rate of 7.2%, both sig- nificantly lower than those of the three comparison models. These results demonstrate that the proposed approach effectively enhances both the privacy and availability of smart meter data while maintain- ing high query accuracy and strong resistance to inference attacks, offering a practical solution for smart power data privacy protection.
Important Links:
Go Back
