[ZEPPELIN-1497] %spark and %python interpreters are able to read credentials.json file

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.6.1
• Assignee: Unassigned
• Reporter: Dmitry Zhukov
• Created: 27/Sep/16 08:29
• Updated: 27/Sep/16 17:14

Description

Currently DB credentials for all users are stored in a plain text file on disk. This file is readable by any interpreter capable of running arbitrary code with Zeppelin user privileges, e.g. %python or %spark.

%spark
scala.io.Source.fromFile("/home/ubuntu/zeppelin-0.6.1-bin-netinst/conf/credentials.json").mkString

res6: String =
{
  "credentialsMap": {
    "anonymous": {
      "userCredentials": {
        "testdb": {
          "username": "user",
          "password": "pass"
        }
      }
    }
  }
}

%python
open("/home/ubuntu/zeppelin-0.6.1-bin-netinst/conf/credentials.json", 'r').read()

'{\n  "credentialsMap": {\n    "anonymous": {\n      "userCredentials": {\n        "testdb": {\n          "username": "user",\n          "password": "pass"\n        }\n      }\n    }\n  }\n}'

Basically if one has %python or %spark interpreters enabled it makes the whole credentials store useless and even dangerous.