Does anyone know how the AppEngine sandbox is implemented? I heard they use a SecurityManager with Java but surely they are using another layer as well like LXC. I've looked at the whitelist and there are a lot of classes that:

  • commonly used in exploit chains because they have security flaws that Oracle refuse to fix (ScriptEngine)
  • had numerous security problems recently and still quite a big unknown (java.lang.invoke.*)
  • had security issues recently and in the past (java.util.concurrent.*, javax.xml.transform.Templates)

I'm interested in running stuff on appengine but if they are only using the SecurityManager then I think they have a huge problem with security.


1 Answer 1


I think it might help with the last part of your question if I point out that the main point of AppEngine security in Python was protecting THEM, not US. I previously read a paper that did have quite a few specifics. At one point, someone specifically said the AppEngine security modifications were to keep customers' scripts from damaging Google's host systems. Stopping malice in general in customer-submitted applications or protecting customer applications from hacker attacks was still an open, unsolved problem.

I'd expect they have the same goals and limitations in the Java runtime.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .