How does it work
the avatar software?
The avatar software, developed by Octopize, is a unique solution for anonymizing personal data. Avatar solves the paradox between protecting individuals' personal data and sharing that data for its informational value. The software, Made in France and successfully evaluated by the CNIL, makes it possible to generate anonymous and statistically relevant synthetic data from the original personal data.
Avatar software offers an ethical value proposition based on a unique combination of benefits:
- Security & privacy
- Reproducibility & Quality
- Proof
- Parameterization
How to use it?
One of our priorities is to make data anonymized easy thanks to avatar data.
With this in mind, we follow 5 main steps to perform an avatarization:
- Original data : Data is imported in a tabular format.
- Data preprocessing : Preferences and constraints are defined.
- Avatarization : Avatar data is generated.
- Postprocessing data: Avatar data is evaluated.
- Avatars: avatar data is ready to be used
The evaluation is based on two pillars: the privacy and the utility.
During the evaluation, several measures are calculated and an automatic report is generated. Compliance is documented and demonstrable. To facilitate the use of the avatar software, the Octopize team offers training, support and documented customers (Python and R).
What the avatar solution allows
The avatar solution is compatible with any type of tabular data, including continuous, categorical data, dates, and geolocations.
The avatar solution includes features applicable on very large data sets and data hierarchical.
How do I install it?
Avatar software can be deployed in a few hours on all infrastructures. We support SaaS and on-premise deployments.
La platform consists of several components: an HTTPS API, a file system to temporarily store datasets, and a database to exclusively contain metadata. The avatar platform can be deployed on a single instance (using docker-compose) or on a Kubernetes cluster. It is generally the data scientists who perform the anonymization via the platform.
The end user (usually the Data Science team) will then install a client library lightweight for the programming language that's right for it (we currently support Python/Pandas, R, and TypeScript). This library simplifies interaction with the avatar API.
Security is one of our pillars: our code is audited using automated static analysis, data is encrypted in transit and at rest. Find more details in our technical documentation.
