[RFE] [RFC] Define way / policy for the expected form of XCCDF rules' names (possible to use interrogative form instead of current's imperative one?)

Shawn Wells shawn at redhat.com
Thu Dec 5 07:10:32 UTC 2013


On 12/4/13, 11:02 AM, Jan Lieskovsky wrote:
> Hello folks,
>
>    as can be seen for example in the scap-workbench demo pictures:
>    [1] https://fedorahosted.org/scap-workbench/attachment/wiki/WikiStart/browsing_results.png
>    [2] https://fedorahosted.org/scap-workbench/attachment/wiki/WikiStart/intro_screenshot.png
>
> the current way how XCCDF rules' names are created is in imperative
> form (e.g. "Set / Ensure / Verify ...") instead of the interrogative
> one (e.g. "Substantive [Subject of the check] meets feature / criteria").
>
> While this approach is natural for HTML version of the guide, the XCCDF
> rule names are used also in tools performing the scan / remediation
> (e.g. scap-workbench:
>    [3] https://fedorahosted.org/scap-workbench/)
>
> And this is where the source of confusion appears / might appear --
> consider the following selected cases (there are far more of them
> than just these though):
>
>            Rule Name                         |  Scan Result
> ----------------------------------------------------------
>   Set Password Minimum Length in login.defs  |  fail
> ----------------------------------------------------------
>   Set Password Minimum Age                   |  fail
> ----------------------------------------------------------
>   Set Password Maximum Age                   |  fail
>
> What the scan is actually doing being comparing if applied system's
> policy for the expected form of passwords:
> * doesn't allow passwords shorter than expected minimum length,
> * if there are minimum / maximum periods how long password
>    (at least / at max) should be considered as valid defined.
>
> But based on the above presented form scap-workbench /
> scap-security-guide user, unfamiliar with the internals,
> might obtain an opinion / impressions when seeing the 'fail'
> result, that the tool actually tried:
> * to set the minimum length for passwords on the system,
> * to set minimum and maximum period during that passwords
>    should be considered valid,
>
> and this act failed (IOW the current form migth induce a perception
> that instead of OVAL check failure [the true reason], the
> remediation failed [red herring / false positive]).

 From the SSG workshops given through N.A., attendees have yet to 
confuse scanning results with remediation state. Historically this 
hasn't been a concern.

With that said, remediation capabilities are rapidly maturing. With 
online remediation becoming available, such as reflected in Šimon 
Lukašík's blog [1], I could see how users would be confused if the 
/scan/ failed, or if the /remediation/ did.

If online remediation is attempted, and fails, the end state of the 
machine is still non-compliance. Perhaps an approach would be to extend 
SCAP Workbench (and OpenSCAP?) reporting functionality to indicate not 
only compliance state, but also if remediation was attempted?


> To avoid this source of confusion we propose to change the way
> how XCCDF rules are named (use interrogative form instead of
> imperative one).
>
> To mention examples for above three cases:
> * 'Set Password Minimum Length in login.defs' would become for
>    example:
>    'User Passwords In login.defs Meet The Minimum Length Requirement /
>     User Passwords Meet The Minimum Length Requirement /
>     Only Passwords Longer Than Defined Minimum Allowed'
>
> * 'Set Password Minimum Age' could become:
>    'Minimum Period Between Password Changes Defined'
>
> * 'Set Password Maximum Age' could be:
>    'Maximum Number of Days a Password Might Be Used Set'
>
> Using this form (we think) the failing scan result wouldn't
> conduct to impressions, the remediation being what failed.
>
> If we could agree on this proposal, I could come with patch
> how Fedora rules (or even RHEL6?) names should be changed
> not to (possibly) lead to this confusion.

There's been discussion on this in the past. Classically the use of SSG 
has enabled (mostly government) users to demonstrate compliance against 
/requirements/. These requirements are imperative, and thus many XCCDF 
titles were mostly written in this form.

A few derived profiles, such as the U.S. DoD STIG, mandate even stronger 
language ("The System shall...", "The system must...."). The ability to 
extend/rename XCCDF titles was accounted for during the U.S. DoD STIG 
process, and reflected in the STIG Overlay (note the <title> tags):
https://git.fedorahosted.org/cgit/scap-security-guide.git/tree/RHEL6/input/auxiliary/stig_overlay.xml

If we migrate XCCDF titles to an interrogative form, we'd need to create 
XCCDF title overlays for not only the STIG, but also USGCB, NIST, FISMA, 
and almost every profile currently existing. Each overlay is some 1,500 
lines long..... which would mean an administrative burden of at least 
another 6,000 lines of code. I'm not sure the effort + burden would be 
worth it.

To clarify OVAL result vs remediation result, perhaps the tools could be 
expanded?


> Comments / feedback appreciated.
>
> Thank you && Regards, Jan.
> --
> Jan iankko Lieskovsky / Red Hat Security Technologies Team
>
> P.S.: Prior actually requesting / proposing this change we
>        have been considering if scap-workbench tool couldn't
>        instead of XCCDF rule names display underlying OVAL
>        check titles (and have these titles customized). But
>        this approach has been realized as not possible to perform
>        because:
>        * there might be more than just one OVAL checks for one
>          XCCDF rule (in that case which one to display?),
Having multiple OVAL mapped to a single XCCDF rule would be... 
challenging. Every XCCDF rule should reflect a unique configuration 
change, preferably to the level of having a CCE assigned.

As an idea, perhaps we should renaming the XCCDF groups to be 
interrogative, with the individual XCCDF rules being imperative?

>        * during parsing the particular XCCDF scap-workbench
>          (some other tools too?) is operating on rule IDs instead
>          of rule names so it would be difficult (if even
>          possible) to modify the code to get it to display
>          OVAL titles instead of XCCDF rule names.
Side note, which may warrant it's own thread: Should we use XSLT to 
insert the XCCDF title automatically into the associated OVAL check? 
This could greatly reduce the maintenance burden if/when any renaming 
occurs.

> P.S.#2: Another possibility which comes to my mind is instead
>          to have XCCDF rules shared in both, the HTML version of the
>          guide and in the validation tool screen, we could dedicate
>          XCCDF rule names to be displayed just in the validation tools,
>          and generate rule names for HTML version of the guide some
>          other way (like keeping it in some other / yet different
>          XCCDF element and when generating HTML guide use content of
>          that element instead of XCCDF rule name).
The maintenance burden of this would be high.


Thanks for starting this thread Jan! It's a very worthwhile conversation.


[1] http://isimluk.livejournal.com/3573.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.fedorahosted.org/pipermail/scap-security-guide/attachments/20131205/283885a0/attachment-0001.html>


More information about the scap-security-guide mailing list