Research On Interactive And Non-interactive Data Publishing Under Differential Privacy

The development of electronic information technology has driven the progress of the digital economy,and the key to the digital economy lies in data.Companies and organizations use big data to summarize experience,discover laws,and predict trends to assist decision-making services,but the process of data collection and release may create the risk of privacy disclosure.Differential privacy technology is applied in the process of data publishing to protect privacy,but it affects the availability of data.This article mainly proposes some improvements to this problem.In the interactive environment,a reasonable privacy budget allocation can lead to higher data availability.In order to solve the problem of privacy budget allocation in limited interactive data queries,this paper evaluates the existing privacy budget allocation algorithm.Then a privacy budget allocation algorithm based on sensitivity is proposed.Theoretical analysis and experimental results show that the algorithm based on sensitivity conforms to ?-differential privacy,and the noise introduced by the sensitivity method is reduced by an average of 4 orders of magnitude compared to the adjusted special series method,which effectively improves the availability of data.In the non-interactive environment,directly publishing differential privacy histograms on streaming data introduces a lot of noise,resulting in low availability of the published streaming data.Based on this,this paper proposes a basic Kalman filter streaming data histogram publishing algorithm,which uses Kalman filtering to postprocess the privacy-protected data in the sliding window.Theoretical analysis and experimental results show that the algorithm is consistent with w-event differential privacy,and the noise introduced during data distribution is 19.2% less than the existing streaming data histogram distribution algorithm.If the sudden change of the streaming data is taken into consideration and the post-processing is performed by the improved Kalman filtering,the noise introduced will be reduced by an average of 28.1 % compared with the existing streaming data histogram publishing algorithm,which improves the accuracy of publishing data.Using the ideas in this article,the availability of data can be guaranteed when querying in the interactive environment,and the high accuracy of streaming data can be guaranteed in the non-interactive environment.These works are conducive to protecting data privacy and allowing data requesters to obtain more effective information in the process of data publishing.