Anonymous collaboration method based on differential privacy under MapReduce framework

Abstract

The invention discloses an anonymous collaboration method based on differential privacy under a MapReduce framework. The method combines an anonymous privacy protection technology with a differentialprivacy protection technology to improve the computation rate under the MapReduce framework; the method is used for solving contradiction between data availability and security, on the basis of the data anonymous privacy protection technology, a Laplace noise mechanism is added to strengthen privacy protection performance, and on the basis that a data set improves the data processing efficiency through the MapReduce framework, it is still maintained that data has high security and availability.

Description

technical field [0001] The invention relates to an anonymous collaboration method based on differential privacy under a MapReduce framework, and belongs to a data privacy protection method in the field of data security. Background technique [0002] In order to process a large number of data sets, traditional data privacy protection technology has encountered a bottleneck, so data parallel computing becomes particularly necessary. Efficient computing can greatly improve the efficiency of data release and shorten the access time of users. [0003] In the field of data privacy protection, the large-scale, high-speed, and diverse characteristics of big data make it different from conventional small data. Micro-aggregation technology can better reduce the information loss rate on the basis of providing data anonymity protection, and the availability is very high. High, but the micro-aggregation privacy protection method has great limitations on big data, mainly because: firstly...

AI Technical Summary

Problems solved by technology

[0003] In the field of data privacy protection, the large-scale, high-speed, and diverse characteristics of big data make it different from conventional small data. Micro-aggregation technology can better reduce the rate of information loss on the basis of providing data anonymity protection, and the availability is very high. High, but the micro-aggregation privacy protection method has great limitations on big data, mainly because: firstly, under large-scale data, it is difficult for unit computing to effectively divide and aggregate data within an acceptable time range, making the processing efficiency Low, secondly, under the diversified data, the background knowledge owned by the attacker increases, and the security of data privacy protection decreases

[0004] In terms of security performance, differential privacy has a more rigorous mathematical-based privacy protection model, but some restrictions on queries after the original data are anonymized can be circumvented by differential privacy that supports a large number of potentially diverse queries. However, differential privacy is guaranteed to modify a The probability distribution of the published data is unchanged when the input record is entered. It is necessary to add necessary noise to the data for disturbance. The larger the amount of data, the greater the noise, which will inevitably cause loss of data availability.

Images