A Differential Privacy High-Dimensional Data Publishing Protection Method Based on Principal Component Analysis Optimization

Abstract

The invention discloses a method for publishing and protecting differentially private high-dimensional data optimized based on principal component analysis, comprising the following steps: step 1, calculating the information entropy of original data attributes, determining the attribute importance threshold, and screening the attributes in the original data ; Step 2, utilize principal component analysis method to carry out dimension reduction to described screening data, determine optimal k value, thereby determine the best release data; Wherein, in the process of dimension reduction, the projection matrix that produces carries out personalized noise addition Obtaining the added data, and making the added data satisfy differential privacy; and in the dimensionality reduction process, performing multiple selections of the number of principal components k, and calculating the relationship between the original data and the The mutual information of the noisy data determines the optimal k value. The present invention provides a differential privacy high-dimensional data release protection method based on principal component analysis optimization, which ensures that data privacy information is not leaked, and at the same time, the released data is better close to the original data.

Description

technical field [0001] The invention relates to the technical field of privacy protection data release, in particular to a differential privacy high-dimensional data release protection method based on principal component analysis optimization. Background technique [0002] At present, many data collection organizations need to release the collected raw data (such as medical data, financial data, etc.) to facilitate data analysis and mining, and to generate more effective decision support from the released data, such as figure 1 Shown is a schematic diagram of data distribution. However, the released original data involves a large amount of personal sensitive information, and directly releasing the data will lead to a serious disclosure of personal privacy. Therefore, data publishers need to process private data through special protection technology before releasing the data. [0003] At this stage, the main privacy-preserving data release technologies are roughly divided i...

AI Technical Summary

Problems solved by technology

However, these data release methods are not suitable for the release of high-dimensional data, and cannot solve the problem of "dimension disaster". The problem

Therefore, how to provide data researchers with a large amount of effective information while using differential privacy technology to ensure the privacy and security of original high-dimensional data has become extremely challenging.

Images