A Method for Protecting the Privacy of Massive Data Based on Differential Privacy

In the context of the demand for large-scale static data release, this paper proposes a security protection method for massive data privacy based on differential privacy. Design an algorithm called Sparkup-Gus (Sparkdi Ferentiar Privasi-Grupinsmuddin), which is protected based on ε differential privacy and is suitable for Spark's memory iteration. The algorithm not only significantly improves the speed of outlier judgment and the efficiency of data difference calculation. It also effectively controls the sensitive influence of outliers in the histogram-based data publishing method on data release. Subsequently, experiments were carried out on the T -Driver trajectory dataset. The results show that compared with other models such as CNN, LSTM and PVTv2, the PVT2-CBAM model proposed in this paper improves the prediction mean absolute error by 45% and 33%, respectively. When the privacy budget is adjusted to 0.3 and 0.7, the average absolute error of the proposed method is increased by about 5% and 28%, respectively, which further verifies the feasibility and effectiveness of the proposed method.