Securing Hudson

In the default configuration, Hudson does not perform any security check. This means any person accessing the website can configure Hudson and jobs, and perform builds. While this configuration is normally acceptable for intranet use, Hudson can be also configured to authenticate users and enforce access control so that it can be exposed to the untrusted environment, such as the internet.

This setting is controlled mainly by two axes:

  1. Security Realm, which determines users and their passwords, as well as what groups the users belong to.
  2. Authorization Strategy, which determines who has access to what.

These two axes are orthogonal, and need to be individually configured. For example, you might choose to use external LDAP or Active Directory as the security realm, and you might choose "everyone full access once logged in" mode for authorization strategy. Or you might choose to let Hudson run its own user database, and perform access control based on the permission/user matrix.

Topics

Security Implication

Note that even when security is enabled, builds that are performed in Hudson still retains the full access to the entire system, because they run as different processes. This implies any committor to any of the projects built in Hudson has full access to the system. They can change the build script to look at any files in the system that Hudson has access to, and modify them.

Also, if you are using the master/slave mode, slaves that are connected to Hudson also gains the full access to the entire Hudson build cluster, since a slave can send code to the master to be executed.

On the other hand, Hudson does not escape any input strings. You can use any HTML tags and also someone can embed harmful scripts. Appropriate security settings must be required.

Labels:

Enter labels to add to this page:
Wait Image 
Looking for a label? Just start typing.
  1. Oct 17, 2008

    David Multer says:

    I understand the security implications on the master, but what is meant by slave...

    I understand the security implications on the master, but what is meant by slaves being able to send code to the master for execution? Can a slave build script specify actions to be executed on the master? I thought distributed builds worked by running the build on a (mostly) replicated copy of the master environment and returning results back to the master. Maybe I lack details on the whole master/slave interaction during a build. More details on this and security will help to assess the best security tradeoffs.