Integrations¶

CSVW-EO metadata can be integrated with Differential Privacy libraries and systems.

Currently supported integrations include:

SmartNoise SQL
OpenDP

SmartNoise SQL Integration¶

The csvw_to_smartnoise_sql.py utility converts CSVW-EO metadata into SmartNoise SQL YAML configuration.

Supported Mappings¶

CSVW-EO	SmartNoise SQL
`maxContributions`	`max_ids`
`minimum`	`lower`
`maximum`	`upper`
`privacyId`	`private_id`

Usage¶

python csvw_to_smartnoise_sql.py \
  --input metadata.json \
  --output snsql_metadata.yaml

Example output¶

table:
  max_ids: 3

  age:
    type: int
    lower: 0
    upper: 100

OpenDP Integration¶

The csvw_to_opendp_context.py utility creates OpenDP contexts from metadata and datasets.

Usage¶

import polars as pl

from csvw_eo.csvw_to_opendp_context import (
    csvw_to_opendp_context,
)

data = pl.scan_csv("data.csv")

context = csvw_to_opendp_context(
    csvw_meta=metadata,
    data=data,
    epsilon=1.0,
)

Design Goal¶

CSVW-EO aims to provide a portable metadata layer between Differential Privacy systems.

The metadata should remain implementation-independent while still enabling automated DP workflows.