An image can not be activated

[Jess Goodman]

(exasperated _sigh_to those who don't understand the concept of security LAYERS.)

Yes, it would be possible for a skilled hacker to get the necessary bits in a file to create an .EXE file anyway. That doesn't mean protecting LINK.EXE is useless.

Hein, Robert - I assume you don not bother to lock your front door. After all, anyone with an axe would be able to get through, so locking it would only give you a false sense of security.

[Jess Goodman]

Hein wrote:
"You may be better of monitoring (audit) linker usage."

Clearly you don't get the "layer" concept. You are not better off using auditing INSTEAD - you are better off using auditing IN ADDITION to all other security measures you can implement, including setting images to no world access that untrusted users have no business running.

[Robert Gezelter]

Jess,

With all due respect, I do lock my front door, and my car for that matter.

However, I do recommend caution on protecting "normal" non-privileged utilities. I have encountered references to them in far too many surprising contexts over the years to want to find additional dependencies as security alarms.

One can also remove things from DCLTABLES, but at the risk that command procedures will suddenly stop working when an otherwise innocuous change is made.

Security can indeed be subtle, and reasonable professionals can disagree on issues like these.

- Bob Gezelter, http://www.rlgsc.com

[John Gillings]

re: Jess and "protecting" LINK.EXE

LINK is an unprivileged image. If I have access to DCL, and a network, or any external media, I can easily get a copy of LINK.EXE from distribution media, or any other system (I'm sure it wouldn't take long to find a copy somewhere on the web). Place the image anywhere I like, point the logical name LINK at it and I'm back in business.

This is nowhere near the domain of "skilled hacker", it's a very basic understanding of how command and image activation work.

It's not like locking the gun cabinet at all, it's more like putting sign on the unlocked cabinet saying "locked". The real problem is the illusion of security where none exists.

The same is true for any image in SYS$SYSTEM that is not required to be installed with privilege to function.

Similarly removing verbs from DCLTABLES and creating a rod for your own back on every update. It's trivially simple to add a command back - even if you also protect the CLDs in SYS$UPDATE (again they're public domain and easy to reconstruct)

[Jess Goodman]

We don't allow our untrusted users to download files from the internet. But even if you do, every extra step you make an attacker take leaves another footprint.

John, you have mentioned in this and other threads that you think it is trivial for a user to put missing commands back into their command tables.

Assuming, of course, that the deleted commands include the SET COMMAND syntax, and the user has the restricted flag, please tell me how to do this trivial act? I can't figure it out.

[John Gillings]

Jess,

> We don't allow our untrusted users to
>download files from the internet

Can they receive mail (even just text)?

>every extra step you make an attacker take
>leaves another footprint.

In this case you'd probably be better off leaving LINK unprotected, but audited. You then know when someone uses it, without tipping your hand. If someone gets a private copy, you don't know when or how it's being used.

>I can't figure it out.

Maybe for the sake of your security I'd better not say ;-)