potiuk commented on issue #33276:
URL: https://github.com/apache/airflow/issues/33276#issuecomment-1672979060

   This is misunderstanding of Airlfow's security model.
   
   DAG author - by uploading a DAG has capabilities to do everything in the 
worker. It is part of our security model and it is not something we are going 
to change, not somethign that is possible to change in the way how bash, Python 
and unix works. If you have a way to read config to perform the task, and have 
a way to write a code that can perform any code, you cannot prevent that somene 
from doing anything - in cluding getting your configuration wrapping it into a 
zip file and sending it to the whole world.
   
   What you are asking for is logically impossible to do. 
   
   This is nicely described in our model. You should read Airflow's Security 
model and understand what capabilities different actors in Airlfow "realm" have 
- we updated and clarified it recently and when we release 2.7.0 it will be 
further updated and clarified:  
https://airflow.apache.org/docs/apache-airflow/stable/security/index.html#airflow-security-model-user-types
   
   It's up to Deployment Manger to set-up the right rules and tooling at teh 
"DAG" submission gates (for example code review, automated checks etc.) if 
Deployment's Manager's intention is to prevent certain actions (it's also 
described in our model)
   
   Also there is an upcoming work under the umbrella of AIP-1 
https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=89066609 and 
"mutli-tenancy" work - which will add a few more things to the model: mainly 
eventually you will be able to limit access to some tenant-specific 
configuration (Variables and Configuration) to only be available for DAGs of 
that tenant, but it does not change the model substantially, it just allows to 
isolate presence of certain sensitive information to only a group of DAG 
authors (but DAG authors in that group will continue to be able to do anything).
   
   You need to adapt your security expectations and your own processes to the 
model, simply. There are no intentions (or even this is actually impossible 
without changing how Python interpreter works and changing some super -basic 
principles on how airflow works) to change it, so my recommendation is not to 
try to wast your time on it.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to