Re: Secure OS Thoughts
From: Douglas A. Gwyn (DAGwyn_at_null.net)
Date: 08/30/03
- Next message: Tao Zhang: "Good paper(book) on system security attacks?"
- Previous message: Roger Schlafly: "Re: Crypto Mini-FAQ"
- In reply to: Martin Bealby: "Secure OS Thoughts"
- Next in thread: Martin Bealby: "Re: Secure OS Thoughts"
- Reply: Martin Bealby: "Re: Secure OS Thoughts"
- Reply: Anne & Lynn Wheeler: "Re: Secure OS Thoughts"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Date: Sat, 30 Aug 2003 02:42:31 -0400
Martin Bealby wrote:
> This set me
> thinking, why has nobody created a secure operating system.
There have been several such.
It is true that it is much more reliable, and more cost-
effective, to build security in from the outset than to
try to retrofit it onto an existing system. What has
been rather well established is that simplicity is a
strong ally when creating secure operation.
The *biggest* single obstacle to establishing security
on the systems most people actually use, is that the
users demand support for all the features (bells and
whistles) that they are now accustomed to, and quite
simply that means using *unsecurable* protocols and
trusting agents of *unknowable* trustworthiness. The
commercial exploitation of the Internet would take a
fatal blow if it were forced to embed genuine security,
and the will and direction necessary for that are just
not present. Even the DoD has surrendered control of
their own communications design and largely gone over
to using crappy stuff like Windows with IPv4. I miss
the good old days when experienced communication
security engineers designed and built our systems.
But it's a lot cheaper to not have decent security,
so...
Excellent security is just *not* a serious priority
and/or the people making such decisions don't have a
clue (same thing in my opinion).
> Needed capabilities:
Yes, but not in the sense that you intended.
Your list was well-intentioned, but neither necessary
nor sufficient to obtain a high level of security, and
it would all be subverted anyway in practice. What is
truly needed is a nonsubvertible underlying mechanism.
Capas (all the way down to instruction level) perhaps.
The IBM System/38 was an interesting experiment that
got *some* of the basics right, but it had no effect on
the design of current sstems.
- Next message: Tao Zhang: "Good paper(book) on system security attacks?"
- Previous message: Roger Schlafly: "Re: Crypto Mini-FAQ"
- In reply to: Martin Bealby: "Secure OS Thoughts"
- Next in thread: Martin Bealby: "Re: Secure OS Thoughts"
- Reply: Martin Bealby: "Re: Secure OS Thoughts"
- Reply: Anne & Lynn Wheeler: "Re: Secure OS Thoughts"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] [ attachment ]
Relevant Pages
|