produceSecureBindings() doesn't seem to work

[asilism]

Hi, recently I found that sandbox can load remote script by using load() function.
It can be secure problem in my project , so I checked option of sandbox.

I found allowLoadFunctions args is "false" by default.

I debugged NashornSandboxImpl, and there is ambiguous part of produceSecureBindings()

private void produceSecureBindings() {
		try {
			final StringBuilder sb = new StringBuilder();
			cached = scriptEngine.getBindings(ScriptContext.ENGINE_SCOPE);
			sanitizeBindings(cached);
			if (!allowExitFunctions) {
				sb.append("var quit=function(){};var exit=function(){};");
			}
			if (!allowPrintFunctions) {
				sb.append("var print=function(){};var echo = function(){};");
			}
			if (!allowReadFunctions) {
				sb.append("var readFully=function(){};").append("var readLine=function(){};");
			}
			if (!allowLoadFunctions) {
				sb.append("var load=function(){};var loadWithNewGlobal=function(){};");
			}
			if (!allowGlobalsObjects) {
				// Max 22nd of Feb 2018: I don't think these are strictly necessary since they
				// are only available in scripting mode
				sb.append("var $ARG=null;var $ENV=null;var $EXEC=null;");
				sb.append("var $OPTIONS=null;var $OUT=null;var $ERR=null;var $EXIT=null;");
			}
			scriptEngine.eval(sb.toString());

			resetEngineBindings();

			engineAsserted.set(true);

		} catch (final Exception e) {
			throw new RuntimeException(e);
		}
	}

It looks like doing method overide such as quit/load/print ... . but after that, call resetEngineBindings() again.
Doesn't it mean that all overrides are initialized?

I don't know if this is the real cause, but in my current sandbox it is still possible to execute load() function.
Is there something I'm missing?

[mxro]

Thank you for raising this issue!

I tried to reproduce it in a unit test:

public class TestIssue134 {

	@Test
	public void test() throws ScriptCPUAbuseException, ScriptException {
		NashornSandbox sandbox = NashornSandboxes.create();

		sandbox.eval("load('classpath:TestIssue134.js')");
		sandbox.eval("load('somethingwrong')");
	}

}

4f3c8fa

It seems like the load function is just the overriden dummy.

When running the same test with the enabled load function as follows:

	@Test
	public void test() throws ScriptCPUAbuseException, ScriptException {
		NashornSandbox sandbox = NashornSandboxes.create();
		sandbox.allowLoadFunctions(true);
		sandbox.eval("load('classpath:TestIssue134.js')");
		sandbox.eval("load('somethingwrong')");
	}

We will get the exception 'I should never be called' from running the script.

Could you help me refine the test so it captures what you have encountered? Thank you!

[asilism]

@Test
	public void test() throws ScriptCPUAbuseException, ScriptException {
		NashornSandbox sandbox = NashornSandboxes.create();
		// create context for Engine scope binding
		ScriptContext context = new SimpleScriptContext();
		// actually, there is more source for binding
		// blah,blah, blah....
		sandbox.eval("load('classpath:TestIssue134.js')", context);
		sandbox.eval("load('somethingwrong')", context);
	}

I run eval() with new sciprt context, but that is not secure context.
Because the context isn't sanitize, and also allow/block function is based on context.
Is there any way to get sanitized context?
Now, I'm confused if this is a glitch or if I'm using sandbox incorrectly. 😂

thanks to every reply.

[mxro]

I think technically the Sandbox worked as expected here.

However, I think your example shows that it is very easy to accidentally expose scripts. Therefore, I released a new version 0.3.0 that attempts to make it safer working with script contexts.

It will now work as follows:

        @Test
	public void test() throws ScriptCPUAbuseException, ScriptException {
		NashornSandbox sandbox = NashornSandboxes.create();
		// create context for Engine scope binding
		ScriptContext context = sandbox.createScriptContext();
		// actually, there is more source for binding
		// blah,blah, blah....
		sandbox.eval("load('classpath:TestIssue134.js')", context);
		sandbox.eval("load('somethingwrong')", context);
	}

Please let me know if this doesn't resolve your issue!