Mono is quickly approaching having a complete implementation of the CoreCLR security model for Mono. This is being developed primarily for use in Moonlight.
This new and simplified security model allows Moonlight to download and execute untrusted code and run it inside a sandbox. A full implementation requires Mono to have an executable image verifier (making sure the binary that we download follows all of the rules and does not try some funny business), an IL verifier that ensures that the code does not contain any unsafe operations and the sandbox system that ensures that the downloaded code only calls methods that it has permission to call.
Click for passable illustration of how the sandbox works.
CoreCLR security can be customized using a handful of attributes. Instead of sprinkling our source code with the attributes and a gazillions #ifdefs we are using our Mono Linker and a few tools and configuration files to reshape our libraries to contain the necessary attributes required to secure the sandbox. We use a number of tools to automate this process and a manual auditing process to audit the results.
This is cool because this is a much simpler sandbox system than CAS ever was and our tools make it very simple for third parties embedding Mono into their applications to create their own sandboxes and reshape what is allowed or not allowed by the sandbox based on their specific needs.
The bad news: this sandbox is only available from trunk right now and will not be easily available until Mono 2.6.