Privacy protection method for associated classification data sequence based on Puffer fish framework

Abstract

The invention relates to a privacy protection method for an associated classification data sequence based on a Puffer fish framework. Firstly, a Puffer fish framework is introduced to formulate a strict privacy protection definition, then, a multi-dimensional Markov chain model is utilized to accurately describe two kinds of relevance between data, and finally, an achievable privacy protection mechanism is proposed to add appropriate noise to protect privacy. According to the method, a privacy protection definition is formulated in a two-dimensional associated data scene based on a puffer fishprivacy protection framework; a multi-dimensional Markov chain model is adopted to reasonably represent two-dimensional relevance between data, a mechanism capable of adding noise is provided in combination with privacy protection definition, and it is guaranteed that the state of each individual at each moment is private data and an attacker cannot distinguish the state while the overall trend of aggregation query analysis is achieved. According to the method, privacy protection of two types of correlation between individuals and correlation in each sequence is considered at the same time, so that the privacy data of the individuals can be protected while the associated data set is available.

Description

technical field [0001] The invention belongs to the field of privacy protection and information security, and more specifically relates to a privacy protection method based on a Pufferfish framework for associated classification data sequences. Background technique [0002] Although differential privacy is a widely used definition of privacy at present, it cannot be applied to the case of data association, because the model assumes that each individual in the data set is independent of each other, which will lead to inconsistency when applying differential privacy directly to the associated data scenario. It cannot meet the privacy definition originally set, resulting in privacy leakage. [0003] Pufferfish's proposal can solve the correlation between data, because it can use the set D to represent the background knowledge owned by all attackers, that is, it can generate all possible probability distributions of the data set, but its shortcoming is that it lacks specific pos...

AI Technical Summary

Problems solved by technology

[0002] Although differential privacy is a widely used definition of privacy at present, it cannot be applied to the case of data association, because the model assumes that each individual in the data set is independent of each other, which will lead to inconsistency when applying differential privacy directly to the associated data scenario. Cannot meet the privacy definition originally set, resulting in privacy leakage

[0003] Pufferfish's proposal can solve the correlation between data, because it can use the set D to represent the background knowledge owned by all attackers, that is, it can generate all possible probability distributions of the data set, but its shortcoming is that it lacks specific possibilities. The implementation mechanism, because all possible probability distributions need to be considered, the computational complexity is too high, and it is difficult to express all distributions

At present, there are some practical mechanisms for specific data sets, but they can only protect the correlation between a single sequence attribute, and cannot be applied to the scenario provided by the present invention

[0004] The disadvantage of the current privacy protection methods for linked data is that they only consider the relationship of one dimension, such as the relationship between individuals or between attributes.

However, many real data sets are composed of multiple related sequences, such as the time series data of different people. The sequences themselves are highly correlated, and the sequences between different people are also correlated with each other. Therefore, the existing privacy methods for linked data Can not be applied to the scene provided by the present invention

Images