Can Anonymeter evaluate risks on pseudonymized or masked data?
Our goal is to create a comprehensive framework that can assess the privacy risks associated with the output of any data protection technique. Anonymeter can currently evaluate privacy risks on synthetic tabular data. However, our development team is working on expanding its capabilities to evaluate the risks of pseudonymized or de-identified data. This will provide organizations with a powerful tool for managing privacy risks across a wide range of data protection methods, ensuring they can protect sensitive data while maintaining compliance with privacy regulations.
Isn’t synthetic data private by default?
The synthetic data generation process irreversibly breaks one-to-one links between synthetic and real data records. This irreversible approach reduces the re-identification risk.
However, the deep learning models used for synthetic data generation might memorize features during the synthesization process. Ultimately, these memorized patterns can be reproduced in the synthetic data, leading to synthetic data privacy leaks.
The risk assessment of synthetic data is left up to each company's discretion. Due to the limited recommendations available, implementing a risk assessment becomes a challenge and puts individuals' privacy at risk. To address this gap, we developed a set of evaluations so you can measure the re-identification risks of synthetic data.
Where can I try Anonymeter?
What types of synthetic data does Anonymeter support?
Anonymeter supports tabular synthetic data.
Can Anonymeter be used for large-scale synthetic data evaluations?
Yes, Anonymeter is designed to be scalable and can be used for large-scale synthetic data evaluations. We have tested it on datasets of several million rows.
How long does it take to receive a privacy report from Anonymeter?
The time it takes to receive a privacy report from Anonymeter depends on the size and complexity of the evaluated synthetic data. However, most evaluations are completed within a few minutes.
Does Anonymeter only measure the re-identification risk of synthetic data?
Anonymeter version 1.0 measures the re-identification risks in synthetic tabular datasets. Anonymeter version 2.0, currently under development, also measures re-identification risks in Statutory Pseudonymized data.
