It's fairly straightforward to design security into a system, which basically comes down to controlling access to memory and the supervisor. The old mainframe security was virtually bulletproof in that regard. The real problem today is that this type of security can be cumbersome to deal with on a highly interactive interrupt-driven OS targeted for workstations as opposed to something intended for servers where the allowed running processes are usually very fixed and restricted. So the security usually gets "detuned" as the saying goes, until people stop complaining. AFAIK there's no easy way to work around that, so design compromises are unavoidable.
If anybody does find a way to get the absolute best of both worlds when it comes to operational transparency and strong security, they should be pleased to know there's a very large fortune and probably a Nobel Prize waiting for them to claim it.