Rasa Allows Remote Code Execution via Remote Model Loading

Vulnerability

A vulnerability has been identified in Rasa Pro and Rasa Open Source that enables an attacker who has the ability to load a maliciously crafted model remotely into a Rasa instance to achieve Remote Code Execution.

The prerequisites for this are:

• The HTTP API must be enabled on the Rasa instance eg with --enable-api. This is not the default configuration.
• For unauthenticated RCE to be exploitable, the user must not have configured any authentication or other security controls recommended in our documentation.
• For authenticated RCE, the attacker must posses a valid authentication token or JWT to interact with the Rasa API.

Fix

We encourage you to upgrade to a version of Rasa that includes a fix. These are:

• Rasa Pro 3.8.18, 3.9.16, 3.10.12
• Rasa Open Source 3.6.21

Once you have upgraded your Rasa Pro or Open Source installation, you will need to retrain your model using the fixed version of Rasa Pro or Open Source. If you have a custom component that inherits from one of the components listed below and modified the persist or load method, make sure to update your code. Please contact us in case you encounter any problems.

Affected components:

• CountVectorFeaturizer
• LexicalSyntacticFeaturizer
• LogisticRegressionClassifier
• SklearnIntentClassifier
• DIETClassifier
• CRFEntityExtractor
• TrackerFeaturizer
• TEDPolicy
• UnexpectedIntentTEDPolicy

If you are unable to upgrade immediately, please follow our mitigation advice below.

Mitigation Advice

• The unauthenticated RCE can be mitigated by ensuring that you enable authentication for the Rasa HTTP API if this is enabled. This means that in addition to passing to Rasa, you should also pass an authentication configuration as described in our documentation .