The differential privacy model has strong privacy protection capabilities. However, there are problems such as low data utility. To protect private data and improve data utility, this paper proposed a differential privacy algorithm based on SOM neural network(SOMDP). First, the algorithm used the SOM neural network model to perform clustering operations on the data, and then added Laplace noise that met differential privacy to each divided cluster. Finally, it theoretically analyzed the feasibility of the algorithm, and evaluated the performance of the SOMDP algorithm on the real data set, the data availability and privacy performance of the algorithm. Experimental results show that, under the premise of meeting the requirements of differential privacy, SOMDP can greatly improve the effectiveness of differential privacy data publishing.

[2021-04-05] Printed Article /doc/2020.05.0137/v2