Personal data detection¶
Availability information: Self-Managed only
This preview feature is only available for US-based Self-Managed AI Platform deployments. Contact your DataRobot representative or administrator for information on enabling the feature.
In some regulated and specific use cases, the use of personal data as a feature in a model is forbidden. DataRobot automates the detection of specific types of personal data to provide a layer of protection against the inadvertent inclusion of this information in a dataset and prevent its usage at modeling and prediction time.
After a dataset is ingested through the AI Catalog, you have the option to check each feature for the presence of personal data. The result is a process that checks every cell in a dataset against patterns that DataRobot has developed for identifying this type of information. If found, a warning message is displayed in the AI Catalog's Info and Profile pages, informing you of the type of personal data detected for each feature and providing sample values to help you make an informed decision on how to move forward. Additionally, DataRobot creates a new feature list—the equivalent of Informative Features but with all features containing any personal data removed. The new list is named Informative Features - Personal Data Removed.
Warning
There is no guarantee that this tool has identified all instances of personal data. It is intended to supplement your own personal data detection controls.
DataRobot currently supports detection of the following fields:
- Email address
- IPv4 address
- US telephone number
- Social security number
To run personal data detection on a dataset in the AI Catalog, go to the Info page click Run Personal Data Detection on the banner that indicates successful dataset publishing:.
If DataRobot detects personal data in the dataset, a warning message displays. Click Details to view more information about the personal data detected; click Dismiss to remove the warning and prevent it from being shown again.
Warnings are also highlighted by column on the Profile tab:
