CVE-2024-35189

Fides is an open-source privacy engineering platform. The Fides webserver has a number of endpoints that retrieve `ConnectionConfiguration` records and their associated `secrets` which _can_ contain sensitive data (e.g. passwords, private keys, etc.). These `secrets` are stored encrypted at rest (in the application database), and the associated endpoints are not meant to expose that sensitive data in plaintext to API clients, as it could be compromising. Fides's developers have available to them a Pydantic field-attribute (`sensitive`) that they can annotate as `True` to indicate that a given secret field should not be exposed via the API. The application has an internal function that uses `sensitive` annotations to mask the sensitive fields with a `"**********"` placeholder value. This vulnerability is due to a bug in that function, which prevented `sensitive` API model fields that were _nested_ below the root-level of a `secrets` object from being masked appropriately. Only the `BigQuery` connection configuration secrets meets these criteria: the secrets schema has a nested sensitive `keyfile_creds.private_key` property that is exposed in plaintext via the APIs. Connection types other than `BigQuery` with sensitive fields at the root-level that are not nested are properly masked with the placeholder and are not affected by this vulnerability. This vulnerability has been patched in Fides version 2.37.0. Users are advised to upgrade to this version or later to secure their systems against this threat. Users are also advised to rotate any Google Cloud secrets used for BigQuery integrations in their Fides deployments. There are no known workarounds for this vulnerability.

Fides is an open-source privacy engineering platform. The Fides webserver has a number of endpoints that retrieve `ConnectionConfiguration` records and their associated `secrets` which _can_ contain sensitive data (e.g. passwords, private keys, etc.). These `secrets` are stored encrypted at rest (in the application database), and the associated endpoints are not meant to expose that sensitive data in plaintext to API clients, as it could be compromising. Fides's developers have available to them a Pydantic field-attribute (`sensitive`) that they can annotate as `True` to indicate that a given secret field should not be exposed via the API. The application has an internal function that uses `sensitive` annotations to mask the sensitive fields with a `"**********"` placeholder value. This vulnerability is due to a bug in that function, which prevented `sensitive` API model fields that were _nested_ below the root-level of a `secrets` object from being masked appropriately. Only the `BigQuery` connection configuration secrets meets these criteria: the secrets schema has a nested sensitive `keyfile_creds.private_key` property that is exposed in plaintext via the APIs. Connection types other than `BigQuery` with sensitive fields at the root-level that are not nested are properly masked with the placeholder and are not affected by this vulnerability. This vulnerability has been patched in Fides version 2.37.0. Users are advised to upgrade to this version or later to secure their systems against this threat. Users are also advised to rotate any Google Cloud secrets used for BigQuery integrations in their Fides deployments. There are no known workarounds for this vulnerability.

The Fides webserver has a number of endpoints that retrieve `ConnectionConfiguration` records and their associated `secrets` which _can_ contain sensitive data (e.g. passwords, private keys, etc.). These `secrets` are stored encrypted at rest (in the application database), and the associated endpoints are not meant to expose that sensitive data in plaintext to API clients, as it could be compromising. Fides's developers have available to them a Pydantic field-attribute (`sensitive`) that they can annotate as `True` to indicate that a given secret field should not be exposed via the API. The application has an internal function that uses `sensitive` annotations to mask the sensitive fields with a `"**********"` placeholder value. This vulnerability is due to a bug in that function, which prevented `sensitive` API model fields that were _nested_ below the root-level of a `secrets` object from being masked appropriately. Only the `BigQuery` connection configuration secrets meets these criteria: the secrets schema has a nested sensitive `keyfile_creds.private_key` property that is exposed in plaintext via the APIs. Connection types other than `BigQuery` with sensitive fields at the root-level that are not nested are properly masked with the placeholder and are not affected by this vulnerability. ### Impact The Google Cloud secrets used for a Fides BigQuery integration may be retrieved in plaintext by any authenticated Admin UI user, except those with the Approver role. Any API users authorized to access the following endpoints may also retrieve the key in plaintext. Endpoints impacted: - `GET /api/v1/connections` - `PATCH /api/v1/connections` - `GET /api/v1/connection/{connection_key}` - `PATCH /api/v1/system/{system_key}/connection` - `GET /api/v1/system/{system_key}` - `GET /api/v1/system/{system_key}/connection` Connection config secret schemas impacted: - `BigQuerySchema` ### Patches The vulnerability has been patched in Fides version `2.37.0`. Users are advised to upgrade to this version or later to secure their systems against this threat. Users are also advised to rotate any Google Cloud secrets used for BigQuery integrations in their Fides deployments: https://cloud.google.com/iam/docs/key-rotation ### Workarounds There are no workarounds. ### Proof of concept Multiple endpoints are impacted, but this PoC will use `GET /api/v1/system/{system_key}` as an example. 1. Using the Admin UI, navigate to `/add-systems`. Add and save a new system `bq_poc`. 2. In the integrations tab of the new system, configure and save a BigQuery integration with secrets. 3. Log in as a different user with any role except Approver and navigate to the `/systems` page. 4. Open the network section of your browser's developer tools. 5. Click on the `bq_poc` system's meatball menu and then click edit. 6. In the network section of browser dev tools you will observe a HTTP GET http://localhost:8080/api/v1/system/bq_poc/ request. In the body of the JSON response the integration secrets values entered in Step 2 are exposed in plaintext i.e. ```json { "secrets": { "keyfile_creds": { "type": "value", "project_id": "value", "private_key_id": "value", "private_key": "value", "client_email": "value", "client_id": "value", "auth_uri": "value", "token_uri": "value", "auth_provider_x509_cert_url": "value", "client_x509_cert_url": "value" } } } ```

The Fides webserver has a number of endpoints that retrieve `ConnectionConfiguration` records and their associated `secrets` which _can_ contain sensitive data (e.g. passwords, private keys, etc.). These `secrets` are stored encrypted at rest (in the application database), and the associated endpoints are not meant to expose that sensitive data in plaintext to API clients, as it could be compromising. Fides's developers have available to them a Pydantic field-attribute (`sensitive`) that they can annotate as `True` to indicate that a given secret field should not be exposed via the API. The application has an internal function that uses `sensitive` annotations to mask the sensitive fields with a `"**********"` placeholder value. This vulnerability is due to a bug in that function, which prevented `sensitive` API model fields that were _nested_ below the root-level of a `secrets` object from being masked appropriately. Only the `BigQuery` connection configuration secrets meets these criteria: the secrets schema has a nested sensitive `keyfile_creds.private_key` property that is exposed in plaintext via the APIs. Connection types other than `BigQuery` with sensitive fields at the root-level that are not nested are properly masked with the placeholder and are not affected by this vulnerability. ### Impact The Google Cloud secrets used for a Fides BigQuery integration may be retrieved in plaintext by any authenticated Admin UI user, except those with the Approver role. Any API users authorized to access the following endpoints may also retrieve the key in plaintext. Endpoints impacted: - `GET /api/v1/connections` - `PATCH /api/v1/connections` - `GET /api/v1/connection/{connection_key}` - `PATCH /api/v1/system/{system_key}/connection` - `GET /api/v1/system/{system_key}` - `GET /api/v1/system/{system_key}/connection` Connection config secret schemas impacted: - `BigQuerySchema` ### Patches The vulnerability has been patched in Fides version `2.37.0`. Users are advised to upgrade to this version or later to secure their systems against this threat. Users are also advised to rotate any Google Cloud secrets used for BigQuery integrations in their Fides deployments: https://cloud.google.com/iam/docs/key-rotation ### Workarounds There are no workarounds. ### Proof of concept Multiple endpoints are impacted, but this PoC will use `GET /api/v1/system/{system_key}` as an example. 1. Using the Admin UI, navigate to `/add-systems`. Add and save a new system `bq_poc`. 2. In the integrations tab of the new system, configure and save a BigQuery integration with secrets. 3. Log in as a different user with any role except Approver and navigate to the `/systems` page. 4. Open the network section of your browser's developer tools. 5. Click on the `bq_poc` system's meatball menu and then click edit. 6. In the network section of browser dev tools you will observe a HTTP GET http://localhost:8080/api/v1/system/bq_poc/ request. In the body of the JSON response the integration secrets values entered in Step 2 are exposed in plaintext i.e. ```json { "secrets": { "keyfile_creds": { "type": "value", "project_id": "value", "private_key_id": "value", "private_key": "value", "client_email": "value", "client_id": "value", "auth_uri": "value", "token_uri": "value", "auth_provider_x509_cert_url": "value", "client_x509_cert_url": "value" } } } ```

Fides es una plataforma de ingeniería de privacidad de código abierto. El servidor web de Fides tiene una serie de endpoints que recuperan registros de "Configuración de conexión" y sus "secretos" asociados que _pueden_ contener datos confidenciales (por ejemplo, contraseñas, claves privadas, etc.). Estos "secretos" se almacenan cifrados en reposo (en la base de datos de la aplicación) y los endpoints asociados no están destinados a exponer esos datos confidenciales en texto sin formato a los clientes API, ya que podrían ser comprometedores. Los desarrolladores de Fides tienen a su disposición un atributo de campo Pydantic ("sensible") que pueden anotar como "Verdadero" para indicar que un campo secreto determinado no debe exponerse a través de la API. La aplicación tiene una función interna que utiliza anotaciones "sensibles" para enmascarar los campos sensibles con un valor de marcador de posición "**********". Esta vulnerabilidad se debe a un error en esa función, que impedía que los campos del modelo API "sensibles" que estaban _anidados_ debajo del nivel raíz de un objeto "secreto" se enmascararan adecuadamente. Solo los secretos de configuración de conexión de "BigQuery" cumplen estos criterios: el esquema de secretos tiene una propiedad sensible anidada "keyfile_creds.private_key" que se expone en texto sin formato a través de las API. Los tipos de conexión distintos de "BigQuery" con campos confidenciales en el nivel raíz que no están anidados se enmascaran correctamente con el marcador de posición y no se ven afectados por esta vulnerabilidad. Esta vulnerabilidad ha sido parcheada en la versión 2.37.0 de Fides. Se recomienda a los usuarios que actualicen a esta versión o posterior para proteger sus sistemas contra esta amenaza. También se recomienda a los usuarios que roten los secretos de Google Cloud utilizados para las integraciones de BigQuery en sus implementaciones de Fides. No se conocen workarounds para esta vulnerabilidad.