ACRO: Tools for the Semi-Automatic Checking of Research Outputs

This repository maintains the ACRO R package, which is an interface to the Python ACRO package.

ACRO is a free and open source tool that supports the semi-automated checking of research outputs (SACRO) for privacy disclosure within secure data environments. SACRO is a framework that applies best-practice principles-based statistical disclosure control (SDC) techniques on-the-fly as researchers conduct their analysis. SACRO is designed to assist human checkers rather than seeking to replace them as with current automated rules-based approaches.

The ACRO package is a lightweight Python tool that sits over well-known analysis tools that produce outputs such as tables, plots, and statistical models. This package adds functionality to:

• automatically identify potentially disclosive outputs against a range of commonly used disclosure tests;
• apply optional disclosure mitigation strategies as requested;
• report reasons for applying SDC;
• and produce simple summary documents trusted research environment staff can use to streamline their workflow and maintain auditable records.

This creates an explicit change in the dynamics so that SDC is something done with researchers rather than to them, and enables more efficient communication with checkers.

A graphical user interface (SACRO-Viewer) supports human checkers by displaying the requested output and results of the checks in an immediately accessible format, highlighting identified issues, potential mitigation options, and tracking decisions made.

Additional programming languages such as this R package are supported by providing front-end packages that interface with the core ACRO Python back-end.

Installation

Install the acro package from CRAN as follows:

install.packages("acro")

Notes for Python 3.13

ACRO currently depends on numpy version 1.x.x for which no pre-compiled wheels are available within pip for Python 3.13. Therefore, in this scenario, numpy must be built from source. This requires the installation of a C++ compiler before pip installing acro.

For Windows, the Microsoft Visual Studio C++ build tools will likely need to be installed first.

If you are unable to install the C++ tools, but are able to install multiple versions of Python, you can use the following method to explicitly create the Python virtual environment with your desired python.exe.

First identify the location of the python.exe needed, which may be in C:\Program Files\ or in your local C:\Users\username\AppData\Local\Programs\Python\.

Then from R, use reticulate to create the virtual environment and install ACRO; see example below. Change the path provided as a version to your Python path (note that forward slashes are required in the path).

>>> library("reticulate")
>>> reticulate::virtualenv_create(envname = "r-acro-0.4.7", version = "C:/Users/username/AppData/Local/Programs/Python/Python310/python.exe", force = TRUE, packages = NULL)
>>> reticulate::py_install("acro", envname = "r-acro-0.4.7")

Finally, the R ACRO library can be used as normal:

>>> library("acro")
>>> acro_init(suppress = TRUE)

Usage

Before using any function from the package, an acro object should be initialised using the following R code:

>>> library("acro")
>>> acro_init()

Then the functions can be called.

Documentation

The github-pages contains pre-built documentation.

Acknowledgement

This work was funded by UK Research and Innovation under Grant Numbers MC_PC_21033 and MC_PC_23006 as part of Phase 1 of the Data and Analytics Research Environments UK (DARE UK) programme, delivered in partnership with Health Data Research UK (HDR UK) and Administrative Data Research UK (ADR UK). The specific projects were Semi-Automatic checking of Research Outputs (SACRO; MC_PC_23006) and Guidelines and Resources for AI Model Access from Trusted Research environments (GRAIMATTER; MC_PC_21033). This project has also been supported by MRC and EPSRC [grant number MR/S010351/1].