Linear regression is one of the important methods of statistical analysis

but the research on linear regression analysis with privacy preserving for multiple data source in distributed environment is relatively less. To solve this problem

by using additive homomorphic encryption

a model on collaborative linear regression analysis with privacy preserving in non-trusted distributed environment is presented. The model makes full use of the calculating ability of distributed environment

the clients and statistical server collaboratively calculate linear regression parameters

and the sensitive data of each data source are protected in analysis process. Finally

the security of model is discussed in Semi- Honest Mode

and the performance of model is tested.