Review Synthetic Data

Synthetic data is artificially generated data designed to reproduce selected characteristics or situations for training, testing, simulation, analysis, or other uses.

Software creates records, images, text, signals, or environments instead of collecting each item directly from the real-world population or process of interest.

Synthetic data can expand rare cases, support testing, or reduce some privacy exposure, but it can also reproduce assumptions, miss important variation, or create hidden feedback loops.

A team generates simulated equipment-failure signals to test detection software against events that are rare in historical records.

Synthetic does not mean anonymous, unbiased, realistic, or safe. Generated records may still reflect real people, source data, or flawed assumptions.

Synthetic data is created to stand in for selected parts of reality.

* [[training-data|Training Data]] * [[data-provenance|Data Provenance]] * [[model-collapse|Model Collapse]] * [[scenario|Scenario]] * [[de-identification-anonymization-risk|De-Identification and Anonymization Risk]]

* [NIST SP 800-188, De-Identifying Government Datasets](https://doi.org/10.6028/NIST.SP.800-188) — authoritative treatment of synthetic data as a privacy-preserving publication technique and its limits. * [NIST, SDNist: Synthetic Data Report Tool](https://www.nist.gov/services-resources/software/sdnist-synthetic-data-report-tool) — official tool and guidance for evaluating synthetic-data utility and privacy.

* Preserve origin and generation method through [[data-provenance|Data Provenance]].