Security Model

Long time listener, first time poster

Can anyone point me to any definite texts on the security model for the
interpreter? I am interesting in understanding the nuts-and-bolts of how
safe levels work.

Cheers!

Mark