We have been using document AI for about a year now doing model development and learning the ins and outs of how these models work.
Currently we are doing a security review as we are about to go to production prototyping.
There was a reported security issue that when running the models, the models were NOT run with the privileges of the initiator, but rather a google service account with excess privileges. I cannot find any additional information on if this has been changed.
Do the models run using the callers privileges or an internal google service account with excessive privileges?
Hi @kkaspszak,
Welcome to Google Cloud Community!
This is an important security concern, and it’s great that you're addressing it before going into production. You’re right to be cautious about a situation where your Document AI models are run with a Google service account that has excessive privileges. This is a valid concern that needs to be clarified, as it could pose a major security risk and violate the principle of least privilege.
Document AI is designed to use the credentials of the caller who initiated the document processing, which should be an end user or a service account specifically for your application.
Here’s a breakdown of possible approaches to ensure you’re not operating with excessive privileges:
Specific Identity Configuration:
Monitor Permissions:
Test:
Contact Support:
Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.
User | Count |
---|---|
2 | |
2 | |
1 | |
1 | |
1 |