Re: Fix Security (was: Paging file)

From: "Leif Svalgaard" <leif@leif.org>
>From: <vhamberg@attbi.com>>
>> So where's the advantage suggested by your remark? To
>> make "bad" changes on the 400, you still have to have
>> got there with sufficient authority in the first place.
>> And you also need to do that in *nix systems. But once
>> in the system, you're "god", or at least halfway there.
>>
>> But is there any way for someone without *ALLOBJ (and
>> probably a few other rights) to do this tampering on a
>> virgin 400? When there's been no program written to put
>> the job into privileged status? Not if your authority is
>> not high enough, I don't think.
>
>You guys just won't see it. On AS/400 number one (yours)
>you have all the authority you need. You make a program
>(normal user state, no adoption of authority) and patch it
>to make it malware. Then recalculate the program validation
>value and save the program. Transfer the save file to
>another AS/400 (the one you are attacking) and restore
>the malware program. For this you need only enough
>authority to restore the program. You might charm the
>system operator or the user to restore it for you. Maybe
>it is a demo program that also does something useful
>and interesting. Once the program is restored (as it
>will with no problems as it is just a user state program;
>you can even sign it for that matter) you run it and now
>you are God on the other AS/400.

Leif is correct about this security exposure. This is why it is a good idea
on V5R2 to set the QFRCCVNRST (force conversion on restore) system value to
6 or 7 to force the retranslation of all restored programs. If the program
was patched this will remove the patches. If the program was created before
V5R1 and does not have observability it will not be restored to the system.
All programs created on V5R1 and V5R2 for the current release will have
enough information for them to be retranslated.

When QFRCCVNRST is set to 6, programs that are signed with a valid
signature from a trusted source are restored without retranslation. So be
careful who you trust.

Ed Fishel,
edfishel@us.ibm.com