Medical autocorrelation time series data differential privacy release method

Abstract

The invention provides a medical autocorrelation time series data differential privacy release method. The method comprises the following steps: firstly, constructing personal medical time series data; dividing the personal medical time series data into a plurality of disjoint and equal-length personal medical time series data sub-sequences through a set sliding window; calculating a normalized autocorrelation function of the sub-sequence according to the sub-sequence, and calculating the periodic sensitivity of the sub-sequence according to a query function; and generating Laplace noise sequences corresponding to the sub-sequences, splicing all the noise sequences to obtain Laplace noise sequences corresponding to the sub-sequences, and adding the Laplace noise sequences to the original time series data to obtain a final to-be-released result. According to the method, the privacy of the user is protected from being leaked to a greater extent, and recently-arrived medical time series data can be processed. According to the invention, on the premise of ensuring that the relevancy of the released data is the same as that of the original data, the security of medical time series dataprivacy protection and the time series data processing efficiency can be effectively improved.

Description

technical field [0001] The invention belongs to the field of medical data protection, and in particular relates to a method for publishing differential privacy of medical autocorrelation time series data. [0002] technical background [0003] Due to the purpose of teaching and scientific research, medical institutions often need to collect and release a large amount of various medical sensitive data. Time-series data in medical datasets refers to a series of data values ​​observed at consecutive time stamps. Using the correlation between data values ​​to analyze and mine time series data can bring huge benefits to the government, enterprises and social public services, such as in disease monitoring, monitoring and analyzing the physical condition of patients can prevent the outbreak of certain diseases . [0004] The above example shows that publishing medical time series data is of great significance to knowledge discovery and acquisition. However, if the data holders di...

AI Technical Summary

Problems solved by technology

However, these models have two shortcomings. First, these models cannot provide sufficient security guarantees, and the models need to be continuously improved for the emergence of new attacks.

Second, these models cannot strictly prove the level of privacy protection, and when the model parameters change, the level of privacy protection cannot be quantitatively analyzed

The length of the noise-added time series data processed in this way is much longer than the original data, and no further processing method is proposed.

Images