Policy creation RFC
By simpler model, I mean (abstractly) one that uses fewer, simpler, or more domain-specific concepts than SELinux RBAC+TE+Identity model. What those concepts are and for what circumstances are they sufficient is what I am working towards now.
Unfortunately, I do not have a specific case or recommendation at this point. Hope to have some ideas soon.
On the other hand, I do expect that such policies will have to be transformed into a SELinux policy for enforcement (i.e., I do not believe that any new things that we really need to enforce cannot be represented by the RBAC+TE+Identity model at present). Also, fixed policies should be expressed in the SELinux model.
Regards,
Trent.
----------------------------------
Trent Jaeger
IBM T.J. Watson Research Center
19 Skyline Drive
Hawthorne, NY 10532
[email protected]
(914) 784-7225, FAX (914) 784-7595
Stephen Smalley <[email protected]> on 05/28/2002 11:24:57 AM
To: Trent Jaeger/Watson/IBM@IBMUS
cc: Frank Mayer <[email protected]>, <[email protected]>, Ian
Sollars <[email protected]>
Subject: RE: Policy creation RFC
Quoted 3 lines · Trent Jaeger
--- Default versus "operational dynamic" policy -- My opinion is that any legitimately modifiable policy needs a simpler model (not just interface) than the SELinux model. As Tom pointed out, the level of expertise of
the
Quoted 3 lines
folks making changes is less than those who established the original
policy. Policies that are likely to be changed by folks with less
expertise require a simpler model (simpler policies will result).
Can you clarify what you mean by a simpler model? By "SELinux model", are you referring to the combined RBAC+TE+Identity security model implemented by the example security server (described in the new policy technical report)? Are you suggesting a simpler model (e.g. Biba rather than TE) or just a fixed base TE configuration?
--
Stephen D. Smalley, NAI Labs
[email protected]