RHEL certified at EAL4+ [LWN.net]

RHEL certified at EAL4+ (versus SLES at EAL4+)

Posted Jun 18, 2007 0:08 UTC (Mon) by smoogen (subscriber, #97) [Link]

In most cases the subprofiles are more important for areas of 'assurance'. CAPP is the 'lowest' level profile and is somewhat equivalent to C2 from the old Orange Book. LSPP is supposed to be equivalent to B1 and the RBAC+LSPP is supposed to move towards B2 (though I have heard differening opinions on this).

http://www.dynamoo.com/orange/orangechart.htm

In any case getting a CAPP/LSPP/RBAC is usually tied to a specific hardware+software combination. I could put the same software on a Dell server and not be 'certified' due to the fact that it requires some hardware dongle thingee that was needed for the protected path on the IBM or HP box.

Getting SuSE to be certified to EAL4+ on LSPP/RBAC is probably underway, but may be much harder to do... Selinux was written pretty much to help meet various EAL high level specifications with all the extra Math ready to be documented when asked for. AppArmour was not and you have to go through a lot of extra hoops to show that it meets the mathematical criteria.

I am not saying it is not impossible.. but that its a lot of extra hoops that in the end SuSE may decide are not a large enough incentive for sales.In the end, EAL certification is not a security standard. One can have a CAPP/LSPP/etc system that any script-kiddie can break into and mutiliate.. EAL is meant to be a validation that if you have a LSPP/RBAC system AND the sysadmin knows what they are doing.. the system should be robust in some sort of environment (which is usually described as being a non-threatening/trusted one.. eg NOT the internet.) Thus in the end having various EAL certs mean that you can be bought/sold to customers who have this requirement listed on their ISO9000 or some other audit list.

RHEL certified at EAL4+ (versus SLES at EAL4+)

Posted Jun 18, 2007 21:08 UTC (Mon) by Wol (guest, #4433) [Link]

If I may quote Einstein:

"As far as the laws of mathematics refer to reality they are not certain; and as far as they are certain they do not refer to reality"

So the maths is no guarantee that things will actually work in the real world...

Cheers,
Wol

RHEL certified at EAL4+ (versus SLES at EAL4+)

Posted Jul 11, 2007 18:34 UTC (Wed) by kreutzm (subscriber, #4700) [Link]

Hello,
as there are some factual incorrect statements, I'll have to correct them.

a) When certifying products you always have to specify the environment. So it depends on the sponsor. Maybe special hardware is required, but I wouldn't expect IBM to sponsor Linux on Dell hardware. Though technically it might behave the same. (Maybe a hardware dongle is required, but I really doubt it)
b) There is no extra math in EAL 4+. You'll need some semi-formal formulation for EAL 5, but formal (mathematical) proofs are only for EAL 6 and EAL 7. But you simply cannot afford it typically.
c) I would consider the PPs the security standard. EAL is only a metric, how much effort went into testing. It is called "evaluation assurance level". The same product may be evaluated to EAL 1, where mainly some documentation is reviewed and up to EAL 7 with mathematical proofs. Just in the first case you are not very sure, that the product does in fact fulfill its promise, while in the latter case you have the mathematical proof.
d) Yes, you always have to read the assumptions. You know, the first certificate for a previous mainstream operating system had networking and graphics turned off ...

But testing is very thourougly, so EAL 4+ is usually not easily breakable. Of course, if the admin usese "123" as root password and ignores all documentation, well ....