Data Stream Aggregation Mechanism Under Local Differential Privacy

In the age of intelligence and big data, the collection and sharing of personal data is becoming more and more common and easy, and with it comes concerns about privacy leakage. How to collect reliable data while protecting user privacy has become an important topic. In order to ensure the accuracy of the results, the privacy budget allocated to each data point should be larger, so the number of data points to be allocated with privacy budget can be reduced. Based on this idea, this paper proposes a data stream aggregation mechanism based on local differential privacy, where the data provider maximizes user privacy by fitting the data stream, perturbing salient points, and ultimately reporting the noisy data, which protects the privacy of the individual data at the source of the data generation; the data collector, after obtaining the noisy data, restores the original data stream through data reconstruction, which provides a data analysis and statistical provides a feasible solution for data analysis and statistics. In order to evaluate the performance and effectiveness of the mechanism, this paper conducts extensive experiments and comparisons on a physical activity monitoring dataset. The experimental results show that the mechanism can well balance privacy protection and data availability, and also achieves a smaller mean square error compared with existing methods, demonstrating better performance.