Open Diffix

What Open Diffix is

Open Diffix is an open-source project that develops tools for anonymizing structured data. Its SynDiffix Python package creates synthetic datasets intended to preserve statistical accuracy while protecting anonymity. It is especially suited to descriptive analysis, including histograms, heatmaps, column correlations, counts, averages, and standard deviations. SynDiffix can generate a separate synthetic table for each analytic task, with anonymity maintained across many column combinations. It can also be used for machine learning, though the maker says it is somewhat less easy to use for those applications. Users can create a distributable blob for generating synthetic tables, but recipients must install SynDiffix, and the maker’s repository says blobs do not scale well beyond about 10 columns. Installation uses pip and requires Python 3.10 or later. When data contains multiple rows per protected entity, the entity identifier must be supplied separately or anonymity may be compromised. Open Diffix also offers pg_diffix, a PostgreSQL extension for anonymized aggregate answers to a very limited subset of SQL queries. Its development is paused; it requires PostgreSQL 13 or higher.