Re: [PATCH] SE-PgSQL/tiny rev.2193

 Greg Stark wrote: 
On Tue, Jul 21, 2009 at 4:24 PM, Joshua Brindle<method@manicmethod.com> wrote: 
You also snipped the other scenario I had where row based access control
isn't required but column level and stored procedure level are.   
Well we already have column level and stored procedure privileges.
 
I understand
you already have column level access controls but it still goes back to how
the user is accessing the data, as a top secret user who can read the column
with full precision or as a secret user with precision removed via a trusted
stored procedure.   
Sure, and people do this every day already with postgres roles and privileges.
 
The SELinux policy would have to give the stored procedure
the ability to read the column and trust it to remove the necessary amount
of precision.   
Well the question is: Is the important feature of SEPostgres the
unification of the privilege model with every other piece of the
system in the SELinux policy? Or is that not the main thing and only
the row-level access security is interesting. None of the use cases
seem to put any emphasis on the unification of the security policy.
 

 I don't understand how you came to this conclusion. Quoting from my prior email:

 ""No, for multiple reasons. First a single person (role) could be logging in at different levels (eg., running the same application as the same linux user with the same credentials) and would need to see different things from the database. The SELinux contexts would provide the differentiation in this case and the SELinux policy would enforce the multilevel policy. ""

 In this case the selinux context (which specifies their MLS level) says the user is running at unclass and there should be no possible credentials or role or anything that gives him access to data above unclass in the database. This _is_ a unification of the policy because whether the unclass user is accessing files or rows in a mixed-level database or entire databases that are classified as such the SELinux policy is governing that. 

 Its the same with selinux aware X. If a user is running 2 database clients, one at unclass and one at secret then each application would how the appropriate results. The policy also disallows that user from copying from the secret database client into the unclass client (in this case there are SELinux rules enforcing behavior of X apps, of the database clients and of the filesystem), unified indeed.