Tom Longfellow wrote:
>1.  Do you really log in to your peripherals that much for it to be an
>issue?   Is this a case of 'We have LDAP, everything must use it'?
>2.  What is wrong with a small self contained local authentication
>method?   No one will stumble across YOU while they are hacking
>your corporate LDAP or AD.

I'm curious if anyone has other answers, but I think there are reasonable 
answers to both of these questions. Employees come and go, and their roles 
change even more often. Thus many organizations understandably want to manage 
identities and roles centrally. Sometimes they're required to do so on pain of 
civil and/or criminal penalties if they don't. It's much easier to manage 
authorizations centrally when they have fewer siloed 
authentication/authorization engines to worry about. It also tends to be easier 
to secure one security engine (or fewer security engines) instead of scores or 
hundreds.

Moreover, a centralized approach is easier for operators and administrators, 
too. It's precisely when you don't log into something too often that you forget 
*that* password precisely when you need it most.

>3.  Security Begins At Home.    What happens when your disk system
>needs a quick adjustment or command to Save your z/OS (or even
>LDAP or Linux) IPL and Recovery? Not staying local can lead you to
>the equivalent of 'locking the keys to your piggy bank INSIDE your piggy
>bank' -- ALSO don't save the only copy of the master FDE encryption
>key inside the disks protected by FDE encryption (piggy bank model 2)

The typical implementation is that there's a "break glass" alternative that 
neatly addresses these concerns. (RACF is analogous.) However, the device in 
question is the IBM TS7700 virtual tape library. In this case it happens to be 
a z/OS-dedicated device but not a z/OS startup device. It seems like you can 
safely assume that (at least one instance of) the IBM Directory Server for z/OS 
(LDAP) will be up, running, and reachable at all times this device is relevant. 
The better analogy is that the TS7700 is the bank, and you can safely assume 
there's always at least some money around (z/OS). Otherwise the bank has no 
relevance.

I would feel more nervous along the lines you suggest about configuring the 
TS7700 to rely on an "external" LDAP server. In such cases you should figure 
out a viable (but still secure) "break glass" alternative if the LDAP server is 
down or unreachable. Note that LDAP servers can synchronize with each other, 
and if the synchronization is interrupted it's not an immediate emergency. 
(Also see below about GKLM.)

>If my key system is in a small, self contained, and properly backed up system,
>I feel better than if I have to go to other organizations and other platforms
>and other networks to support the basic functions of the device.

"It depends." Security is sometimes hard, and there are reasonable arguments 
for having a dedicated, centralized security operations team handling various 
housekeeping functions. For example, if you forget to renew a certificate then 
your organization can have a really bad day. Maybe you "forgot" because you're 
stuck in a hospital ICU? A competent, centralized security operations team 
would be less likely to forget.

I think here the decision was made to centralize certain security functions to 
some degree (TS7700 to z/OS LDAP) but not completely (to some other LDAP 
server). Although these LDAP servers can replicate with each other, so if 
desired there can be logical centralization without excessive physical 
centralization. That's an attempt to balance competing interests, and I think 
it's a good one actually. I've recommended exactly this sort of approach in 
other situations (replicated LDAP servers with z/OS hosting one logical 
instance).

>I was recently forced to move my SKLM key servers from hardware
>under my control to Virtual Machines that they Promise will be
>available.   I did bring up the point that one screwup  that is now
>out of my control will DESTROY the ability to open the DS8K data
>arrays. I was so happy to find that SKLM functionality became an
>internal feature on USB sticks, BUT the hardware they purchased
>still does it the old way.

I believe we suggest that you deploy IBM Security Guardium Key Lifecycle 
Manager (GKLM) in at least a couple "places." For example, you can run GKLM in 
the z/OS Container Extensions *and* on VMware. Or you could run GKLM in the 
z/OS Container Extensions "on premises" (on one or more of your machines) and 
also in a public commercial cloud such as IBM Cloud Hyper Protect. We do offer 
a USB-based alternative, true, but then you have some other management and 
security considerations. That's why there are multiple options.

Your IBM DS8000 series storage units already have some external dependencies, 
and you presumably take reasonable steps to assure that those external 
dependencies are well implemented and managed so that they don't jeopardize the 
availability and performance of your storage units. External dependencies 
include cables, electrical power, and (often) SAN switches. Is your key manager 
any different conceptually? No, not actually. So just have "a couple," as you 
do with redundant cables, power, and switches. And if you're nervous about 
putting all GKLM instances in one type of deployment environment — I can 
understand that — then just spread GKLM across two environments. zCX is an 
excellent one.

— — — — —
Timothy Sipples
Senior Architect
Digital Assets, Industry Solutions, and Cybersecurity
IBM zSystems/LinuxONE, Asia-Pacific
[email protected]


----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to