[RFC] Manual Profile Use

Kevin Spargur kspargur at redhat.com
Tue Aug 14 03:04:51 UTC 2012


On 08/13/2012 10:47 PM, Shawn Wells wrote:
> On 8/13/12 7:35 PM, Spencer R. Shimko wrote:
>> On 8/13/12 10:37 AM, "Jeffrey Blank" <blank at eclipse.ncsc.mil> wrote:
>>
>>> I don't quite follow.  Is this a list of Rules for which no automated
>>> checking is possible?  Or is it a list of things whose remediation
>>> cannot be automated?
>>>
>>> (Essentially, will this be the set (or subset) of all Rules that cannot
>>> be expressed in OVAL?)  (Admittedly there are a few that may be
>>> expressible in the SCE/scripts, but let us avoid that conversation for
>>> now.)
>>>
>>> This is a big topic now.  In transition_notes.xml, you will see a 
>>> <note>
>>> with a list of references from the RHEL 5 STIG which are policy/manual
>>> checks; we are in the process of determining for the STIG profile (once
>>> we understand fully what a STIG should be) whether these 
>>> non-automatable
>>> checks should be included.
>> I do believe Michael's suggestion is combination of those you listed 
>> above
>> plus those related to remediation.
>>
>> There is a large set of SSG content that is process-specific and 
>> automated
>> checks can not be developed.  To my knowledge those are the ones 
>> currently
>> be discussed in transition_notes.  However, there are some that can 
>> not be
>> reasonably "assured" (sorry I hate that word) via the existing OVAL
>> checks.  Some of the existing logic is based on the premise that the
>> system in question was generated via a default RHEL config. But much of
>> it relies on the fact that the state of the system was *always* in the
>> state found when the content was interpreted.
>>
>> For example, password complexity - yes OVAL can ensure the current
>> configuration is correct.  However, it fails to ensure the existing
>> passwords met the specified requirement.  The PAM configuration used 
>> when
>> the password is created is not stored in /etc/shadow. A root password or
>> user password entered during Anaconda install may not meet the 
>> requirement
>> thus the results of the SSG post-install content interpretation are 
>> wrong.
>>   I might be missing something there, but many of the assumptions are 
>> based
>> on the fact that the system being audited via SSG content was *installed
>> and booted* w/ the relevant configuration.  Building on that 
>> assumption is
>> the fact that the configuration was laid down prior to a reboot and the
>> audit was run after reboot with *no* configuration changes between those
>> two (e.g. Ensure ntp is enabled, reboot, thus it is enabled).
>>
>> In addition to that set of content (process-related + not validatable
>> through automation) there is a set of requirements that can be verified
>> through automated means but can not be met through automated means.
>> However, I have a feeling the correlation between automated auditing and
>> automated remediation sets is strong.   Checking for a 12 character
>> password is not sufficient.  Ensuring the system is configured properly
>> *and* prompting for a password addresses that issue.
>>
>> Undoubtedly you guys have put more thought into the SCAP content than I
>> have.  But this strikes me as a rather systemic issue that must be
>> addressed to ensure interpretation of the SSG content results in an
>> accurate report.
>
> Would an actual profile be called for, or just identification of rules 
> that need to be manually validated?
>
>> Yes Michael and myself would like to submit a manual check profile to 
>> SSG.
>>   It would address both situations.  If this isn't acceptable we are 
>> quite
>> willing to carry a patch that addresses these issues.
> So you want to carry a Tresys fork of SSG?   o_O
> I'd say toss the patch over so we can see exactly what you're talking 
> about.
> _______________________________________________
> scap-security-guide mailing list
> scap-security-guide at lists.fedorahosted.org
> https://lists.fedorahosted.org/mailman/listinfo/scap-security-guide

If your trying to ease the effort on the part of neighbor projects such 
as aqueduct or secstate, wouldn't it make more sense to have a automated 
only profile?  Perhaps I've misunderstood your use case.

The transition notes contain items that are manual and automated. There 
are certainly some items that can't be tested for in an automated 
fashion which should be addressed in a manual process if need be.  I 
don't disagree with your assessment that checking the running values as 
well as the configuration should be done where possible.  Patches 
welcome :)

Cheers,

Kevin


More information about the scap-security-guide mailing list