Cloudera Data Science Workbench connects to your cluster like any other client. It relies on your cluster’s built-in security layer to enforce strong authentication and authorization for connections from Cloudera Data Science Workbench.
Cloudera Data Science Workbench users attempting to connect to a secure CDH or HDP cluster must first authenticate themselves using their Kerberos. Once a user has been successfully authenticated, Cloudera Data Science Workbench then depends on your cluster’s built-in authorization tools (such as HDFS ACLs, Sentry, Ranger, and so on) to enforce their own existing access control rules when a Cloudera Data Science Workbench user attempts to access data on the cluster.
For example, let’s assume you are running a secure CDH cluster with
Impala (secured by Sentry). On this cluster, only the
impala-admin group has been granted the required
Sentry privileges to modify data in Impala. If a Cloudera Data Science
Workbench user runs a Python job that attempts to modify data in an
Impala table, Sentry privileges will be enforced. That is, if the user
is not a member of the
impala-admin group, the access
request will be denied.
Similarly, if a Cloudera Data Science Workbench user is attempting to submit Spark jobs on secure HDP clusters, Ranger policies (via the respective HDFS and YARN plugins) will be enforced to decide whether the user has the permission to access the requested directories in HDFS, and whether they can submit jobs to the Spark queue.
This means as long as your cluster components have been secured with access control rules (via ACLs or Sentry or Ranger) that forbid specific users/groups from performing certain actions on the cluster, then all access attempts from Cloudera Data Science Workbench users will also be subject to those rules.