Heteroscedasticity difference privacy preservation method for medical data based on OPTICS clustering

Abstract

A heteroscedasticity difference privacy preservation method for medical data based on OPTICS clustering is proposed. The time complexity of OPTICS clustering algorithm is reduced by introducing singlelinked list update and pointer S, and the combination of K-anonymity and differential privacy preservation enhances its security. In order to ensure the availability of data, In this process, heteroscedastic noise is adopted to improve data availability. During the process, we assume that the attacker can obtain the probability of obtaining the privacy information successfully under the maximum knowledge background, and set the upper bound of privacy parameters so as to ensure that the relationship between data availability and privacy security is effectively balanced within the scope of privacy protection.

Description

technical field [0001] The invention relates to the technical field of medical data protection, in particular to an OPTICS clustering-based privacy protection method for medical data heteroscedasticity difference. Background technique [0002] In recent years, the combination of medical data and artificial intelligence has made intelligent medical treatment unprecedentedly hot. The correct use of medical data has brought great value. Coupled with the combination of artificial intelligence, it has led the medical industry to a new way, new vision, and new situation. , however, due to the value of medical data, the number and methods of attacks on medical data have increased significantly, and the leakage of medical data privacy has become more serious. How to ensure the availability of medical data while ensuring its data Privacy security has become a major focus. [0003] Data Privacy Protection: [0004] In order to achieve data privacy protection, the current common tech...

AI Technical Summary

Problems solved by technology

[0004] In order to achieve data privacy protection, the current common technologies can be divided into three categories: data encryption, restriction of release and data distortion. Although the above methods are very effective, they also have their shortcomings: data encryption makes data availability extremely low, which violates The main premise of the restriction release does not define a strict attack model. When the attacker has sufficient knowledge background, the protection given by this technology is very weak; the typical technology corresponding to data distortion is differential privacy protection technology, which is also the most effective and The most commonly used technique, which defines a powerful attack model, but because it is based on distortion, it will still lead to a significant decrease in data availability when giving data strong protection (increasing noise)

[0006] The OPTICS algorithm is a typical clustering algorithm in artificial intelligence. It can cluster according to the shape of the dense area covered by the data set, find clusters of any shape, and it is not sensitive to the input of parameters, which better reflects the query The density relationship of the data in the data set enables more effective analysis of the distribution and association of medical data, but its processing of low-frequency data needs to be improved