DataSHIELD employs a client-server architecture where the data remains on the server side, typically within the institution that owns the data. Researchers, acting as clients, send commands to the server to perform specific analyses. The server executes these commands and returns the results to the client. This process ensures that the raw data never leaves the server, thereby maintaining its confidentiality.
The analytical commands are designed to be privacy-preserving. For instance, they may involve summary statistics, regression models, or other statistical techniques that do not require direct access to individual-level data. This way, sensitive information is not exposed or transferred between institutions.
