Thread: Java Security Implementation for Plugin Supported Architecture

Here's a small starting point, maybe:
I have a small app that uses a Security Manger to prevent the programs it executes from doing a System.exit(0)
It uses this call to set the Security manager:
System.setSecurityManager(new MySecurityManager());

Ok. So you can set your own security manager, which I am assuming is a subclass of SecurityManager, but how do you set permissions for it?

Do you override checkPermission(...) so that it throws an exception under your terms or do you actually set a Permissions object?

I extended SecurityManager and overrode the checkExit() method for my application.
It throws an exception:
throw new SecurityException("ExecuteJava - ignoring exit!");

Ah ok I see. But how do you know it isn't your code calling for the exit? Do you name your threads or something like that?

Write the code, execute it and see what happens. See what values are available that will tell you what you want to know.

The only parameter is the exit code. I don't understand how that tells me what I wanted to know...

I made the subclass, overrode the checkExit(int) method and called System.exit(0) from another class. I put a println in the overridden checkExit() method and it printed the exit value. That's it.

What if your code uses a "secret" exit code?

I checked Sun's source code. Their implementation of checkExit(int) calls checkPermission(new RuntimePermission("exitVM."+status))

Originally Posted by Norm

What if your code uses a "secret" exit code?

Ah ok I see what you're saying. But... isn't that bad practice since the exit code reports to the underlying OS if the exit was normal or not?

Can the checkExit method change the value being issued by the master app?

Aha! Good idea!

But... say you do want to use the actual java security Permissions classes... how do you use them? I can't find a list of property values anywhere in the documentation.

If your application is open source, secret codes won't do you much good unless you're randomly generating them.

My app is for my use. It's for executing a number of my other apps on a single JVM. All the other apps exit with 0 so no problem. I wanted to be able to execute any of my apps with this one program without any mods to them.

Sun's source code in Runtime's exit method:

    public void exit(int status) {
        SecurityManager security = System.getSecurityManager();
        if (security != null) {
            security.checkExit(status);
        }
        Shutdown.exit(status);
    }

That means whether the SecurityManager changes the exit status is irrelevant. The original call will be used either way, unless you override Runtime, which wouldn't work because the JVM constructs its own Runtime object.

It's no problem for my code.

Ok.

Have you ever worked with the Java security framework much before? I just can't make sense of the policies and such...

No, I don't think I have.
Other that changes to the .java.policy file.

I'm starting to feel like its going to be really difficult to figure out how to handle what I need using the policy files.

I might just design my own encrypted policy file using the ideas you stated here.

Secret exit code and such can just go in the encrypted policy file. Read it... check it in the manager... that might work.

Can the program generate a random "secret" exit code on start up instead of hardcoding it?

Yes that is possible. I've been trying to use the a method where I override the SecurityManager methods I want to restrict and then just put code in each one that says if the current thread is a member of the PluginThreads ThreadGroup, block the action. This worked.... up until I started adding more overrides. Then I started having problems with Sun's internal code throwing SecurityExceptions when doing something as simple as displaying a window...