*Generalized* Framework for Access Control (By Abrams&Lapadula)

GFAC is quite simple. It has one goal: Have a way to run all the security models or concepts you want at the same time.

Of course, you've to be secure. This implies that GFAC should:

1. Check accesses for every security model you have
2. Return a simple decision (Allow, Deny, Do not care)

GFAC therefore allows:

1. Easily extendable: just add a new security model as module.

In details, with RSBAC:

Every sensitive access has a “hook in the kernel”. In other words, a hook to system resources and capacities. The hook, also called Access Enforcement Facility (AEF) asks if it should allow the access or not. Thoses accesses are like: Read a file, Write to it, Append something, Establish a connection, Execute a program, Asks a Linux capacity (raw socket, …), Switch userid, …

This hook, the AEF, asks the Access Decision Facility (ADF) what to do. The ADF asks every security module, which will check if it grants the access, denies, or do not care about it. If any module denies the access, ADF will reply that the decision is to deny the access, else it will grant the access.

A small graph explains things better and in more details, though.

I know its ugly!!