Many applications have the need of being extensible either through an add-in or a scripting mechanism. Technically both can be implemented very easy using the .NET framework at all:
· Supporting add-ins often means nothing else than supplying an interface that have to be implemented by an add-in and loading the add-in at runtime through Assembly.CreateInstance(type…).
But very often as enthusiastic developers implementing this cool add-in or scripting mechanism we forget about security at all (or say, the danger of forgetting is highJ). How can we ensure security in that case? Exactly that’s what I want to shortly right now.
Basically, the answer is quite easy. Just load the assembly and types of the add-in (or the compiled script) into a separate application domain and restrict the security policy of this domain to a minimum of required permissions. Take a look at the following code snippet:
private void SetDomainPolicy(ref AppDomain scriptDomain)
PolicyLevel level = PolicyLevel.CreateAppDomainLevel();
// Create a new permission set and add the permission set to the policy level
PermissionSet scriptSet = new PermissionSet(PermissionState.None);
// Add a policy statement
PolicyStatement stmt = new PolicyStatement(scriptSet);
CodeGroup group = new UnionCodeGroup(new AllMembershipCondition(), stmt);
level.RootCodeGroup = group;
// Apply the policy level
This function just gets a reference to the application domain previously created through AppDomain.CreateDomain and sets the code access security policies to this application domain manually. After you have set those policies through SetAppDomainPolicy you can load new types and assemblies into the app domain running with the restricted set of permissions.
Another important point here is that the add-in or script is not able to perform critical actions by itself – it is the task of the host application to provide an appropriate object model to the add-in / script through which it can do call-backs into the host application for performing some critical actions. Therefore the host application perhaps has to perform some asserts on permissions the add-in / script does not have to avoid leading up in security exceptions (remember: stack walk and don’t forget to revert all asserts in your host application). You can find examples for assert on our local community web page www.dotnetexperts.at within the article for demos-download of our security road show.
What I have found very interesting with this is the existence of a whole object model for code access security. You even can refer to configured code groups and permission sets with this object model or create new ones without touching the .NET Framework Security Tool or caspol.exe anyway. I think I will focus on exactly this object model with my following blog entries…
You can find a detailed description of the things I have described above in the book “.NET Framework Security” which is in my opinion a very good reference but not a book for reading from the beginning to the end at once.