T.R | Title | User | Personal Name | Date | Lines |
---|
231.1 | What issues are the auditors paranoid about? | SIOG::T_REDMOND | Thoughts of an Idle Mind | Sat Mar 14 1992 14:38 | 10 |
| I doubt very much whether there are any articles outlining security
procedures within ALL-IN-1. What specific details is the customer
interested in? ALL-IN-1 is, after all, an application running under
VMS, so it's the normal VMS system security that forms the cornerstone
of any security measures that ALL-IN-1 requires. It's worth noting at
this point that ALL-IN-1 V3.0 makes much better use of things like VMS
identifiers and rights, and that captive accounts have been made fairly
watertight.
Tony
|
231.2 | Auditors like audit trails | IOSG::SHOVE | Dave Shove -- REO-D/3C | Fri Mar 20 1992 14:28 | 5 |
| Also, v3.0 will (optionally) record all system admin operations which
affect accounts (creation, deletion, etc). Auditors tend to like this
kind of thing, which is why we added the feature.
D.
|
231.3 | ALL-IN-1 Network Security | THEBAY::SWANDBY | | Wed May 20 1992 02:52 | 39 |
| Hi --
I've been asked to do a security evaluation on an ALL-IN-1
installation, and I've been out of the ALL-IN-1 arena for about two or
three years. The type of evaluation is *very* specific -- they're
concerned about the ability of captive users to access in any way
(either retrieving or depositing data, files, etc.) on another node in
the network which does not run ALL-IN-1. Mailbus/Message Router is not
used on either system. A further complication is that the non-ALL-IN-1
system has received a poor rating on its security audit -- it's a
pretty wide open system. The auditors are loathe to grant us network
access to it unless we can demonstrate that we've prevented our users
from messing with it over the network. (Yes, we suffer for the error
of their ways.)
So I'm interested in finding out any ALL-IN-1 functions that could have
any impact over a network on a non-ALL-IN-1 system. Obviously,
Document Transfer and Electronic Mail (via SPECIAL.COM) come to mind.
What others might there be?
I plan to review all the functions, but my customer needs buyoff from
their internal auditors before they can go into production -- cutover
is scheduled for June 1. You can see that time is of the essence so
your input could help me meet that deadline.
The customer is on V2.4 and does not intend to upgrade in the immediate
future so if there are measures we should take to plug holes in captive
user accounts, I'd appreciate pointers on them, too.
Finally, they're using integrated WordPerfect. I'm not at all familiar
with its features. My customer has told me that it has an include-file
functionality that could potentially pull a file over the network.
Does anyone know if this is true -- and if it is, if it would be simple
or hard to trap such requests and head them off at the pass.
Thanks much for your input,
Joel
|
231.4 | Go to V3 for a much more secure time | BUFFER::VICKERS | Rearranging the DEChairs | Wed May 20 1992 04:04 | 17 |
| Would it make sense to not give the ALL-IN-1 users the NETMBX
privilege? It doesn't sound like they would need it.
Security is very much tightened in ALL-IN-1 V3.0 over V2.4 so moving to
V3.0 would be a very good idea. We probably shouldn't discuss too many
of the gory details here as policy is to keep security discussions as
vague as possible so as to not spread the knowledge of 'features' too
far and wide. (The moderators will appreciate all the gentle writers
in this conference from debating this point here, thanks.)
The most visible area of improvement is in keeping users from executing
their own code and getting to DCL and ALL-IN-1 functions. There are a
few tools available in DECUS to hack in some crude checks for this in
V2.x but V3 is the best choice, by far.
Good luck,
don
|
231.5 | Well, gee.... | IOSG::PYE | Graham - ALL-IN-1 Sorcerer's Apprentice | Wed May 20 1992 13:49 | 22 |
| So help me here...
These people have got a very insecure system, and they want you to
protect their system from being sc***ed up!!
Seriously (??) what sort of legal access are yoou going to do to this
machine (perhaps I should start calling it a leaky sieve??!) and what
sort of access do they want to stop you doing?
Obviously you can DT SV across the network if the remote system is
insecure. GOLD/Write out of the editor and Print to File gives you the
chance to do this too. Simon says there are some ways to do it from CM
too.
You also need to take the steps that Don said we wouldn't talk about in
.-1 to stop them getting to DCL too.
Not feeling very positive about this,
Yours sympathetically,
Graham
|
231.6 | I know the customer's always right, but . . . | IOSG::SHOVE | Dave Shove -- REO-D/3C | Wed May 20 1992 14:21 | 6 |
| I agree with Graham - if they're worried about access to their insecure
system, they should improve its security (it's easy enough to configure
a system to disallow ALL remote file access to it, if required - this
is a DECnet setup thing).
D.
|
231.7 | It's their error but our problem | THEBAY::SWANDBY | | Wed May 20 1992 18:32 | 23 |
| You're absolutely right -- the owner of the other VAX should secure it.
They're in the process of doing that (or so they tell me -- I only came
on board yesterday). The problem here is time frame. My group needs
to come on line June 1; the other VAX has production data which we need
to access through sanctioned channels; the position of the internal
auditors is that letting us on the net with the other system before they
clean up their act just provides one more hole through which the
integrity of that system can leak. The flag was raised here because one
of our power users used Document Transfer to suck over something he was
supposed to get thru bureaucratic channels. So we need to reassure
them that we've prevented our users from doing such things in the
future if we want to interface with their machine.
If need be, we'll just shut down the network on our side except for a
file transfer our application needs to do in the middle of the night.
We're just trying to avoid such a sledge-hammer approach because our
users do like to do things like send mail to VMS mail users on the
other side.
Anyway, thanks for the input thus far. If anything else occurs to you,
I'm open to hearing it.
Joel
|
231.8 | Call me confused | HOTAIR::MADDOX | When in doubt, change it | Wed May 20 1992 20:16 | 36 |
| From .3
> concerned about the ability of captive users to access in any way
> (either retrieving or depositing data, files, etc.) on another node
> in the network which does not run ALL-IN-1.
From .7
> The problem here is time frame. My group needs to come on line June
> 1; the other VAX has production data which we need to access through
> sanctioned channels;
Which will it be. Do you want users to be able to access data across the
network or not. If the answer is, "there is some data we want users
to access and other data we don't want them to access," (which is what
it sounds like to me) then the decision will have to be made and enforced
at the source of the data (i.e., file protection).
Also from .7
> We're just trying to avoid such a sledge-hammer approach because our
> users do like to do things like send mail to VMS mail users on the
> other side.
They're concerned about users' being able to deposit files on the other
machine but want to preserve the ability to send mail to that machine?
What is mail if it is not a file? Again, if there are certain directories
into which the users should not be able to deposit something, then these
directories will need to be protected.
I can't imagine how ALL-IN-1 (or anyother application) on one machine
can compensate for the lack of security on another system.
With this and 12� other opinions you can buy a cup of coffee.
Joe
|
231.9 | "Should" .nes. "Is" | THEBAY::SWANDBY | | Thu May 21 1992 01:03 | 40 |
| re .8
> Which will it be. Do you want users to be able to access data across the
> network or not. If the answer is, "there is some data we want users
> to access and other data we don't want them to access," (which is what
> it sounds like to me) then the decision will have to be made and enforced
> at the source of the data (i.e., file protection).
Yes, the ultimate goal is to allow users access to some data and not to
other. Howver, the customer is somewhat naive technically regarding
ALL-IN-1 and would like a laundry list of all the ways users could
touch another machine over the network. Then they will go through the
list and say, "Yes, this function is fine -- no threat there," or "No,
we don't want users to be able to do that."
> They're concerned about users' being able to deposit files on the other
> machine but want to preserve the ability to send mail to that machine?
> What is mail if it is not a file?
Well, yes, but there is a difference between someone sending VMS mail
to another user and someone depositing bogus data where it shouldn't
be. Those distinctions are part of the review to take place.
> Again, if there are certain directories into which the users should not
> be able to deposit something, then these directories will need to be
> protected.
Everyone agrees the other system needs to clean up its act including
the owners of the other system. However, that's not going to happen
before our deadline -- it is not in our control. We do control how we
allow our users to relate to that machine. It doesn't strike me as
unreasonable to try to block users on our end from doing particular
functions that are deemed undesirable in the present environment while
allowing those that are not. In a perfect world, we wouldn't even have
to consider this, but it's not perfect here so I'm trying to help them
with a workaround. I hope we can avoid going down a rathole of what
the other system should do -- because you're right so there's no need
to argue the point -- it's just not reality today.
Thanks, Joel
|
231.10 | Where is your time most usefully employed? | IOSG::PYE | Graham - ALL-IN-1 Sorcerer's Apprentice | Thu May 21 1992 09:01 | 10 |
| I would guess that you'd be able to log onto the other system and fix
up its security in less time than it would take you to do all the
necessary customisations to ALL-IN-1 (and local VMS!) to stop the
remote access.
IM(H)O,
Graham
PS Saw your mail, I'll get to it later...
|
231.11 | Some ideas | IOSG::TALLETT | Making broccoli | Thu May 21 1992 09:39 | 31 |
|
Come on guys, get real, we can all see where this SHOULD be done,
but that is not what this caller is asking for, so lets stop
discussing it.
I think the idea of removing NETMBX is the right general direction
on this one. Trying to think of all the possible ways a user could
access the net is too hard - you're going to miss one. If the users
need SOME NETMBX, then maybe you could install the bits that need
it with the NETMBX priv.
Obviously if you want to connect this system to the net then
removing NETMBX from all your users is no good, otherwise why
connect it in the first place?!
Another idea would be to have captive accounts and customise off
"dangerous" options, but this is more error prone as you could,
for example, access the net from within an editor. You could maybe
remove NETMBX from all accounts except one, and have that account
set up to do JUST the thingy that needs the NETMBX.
If the data that you want to access can be copied to your system,
how about getting the other system (the insecure one) to "PUSH"
the data onto some area on your system at regular intervals, and
then your users don't need NETMBX to access it?
A user without NETMBX cannot access the net, except by using an
image installed with privs - its fairly bullet proof.
Regards,
Paul
|
231.12 | Being creative . . . | IOSG::SHOVE | Dave Shove -- REO-D/3C | Thu May 21 1992 12:52 | 15 |
| Paul's suggestion of removing NETMBX from the user accounts is a good
one. However, the users will thereby lose the ability to send mail
EXPRESS (as the Send operation does a DECnet connection to the local
Message Router). Normal (First_Class) mail will be OK (as long as you
don't remove NETMBX from the ALLIN1 account!)
There are an awful lot of places where a user can refer to a VMS file,
for example - DT RV, DT SV, Gold-Get in the editor, etc etc. It would
be hard work (but possible) to customise those options to reject any
filespec that contained a double colon. That still wouldn't prevent a
user with DCL access from assigning a logical to point to a remote
file, but if they can get DCL access they can use DCL COPY etc to
access the remote files anyway!
Dave.
|
231.13 | My last attempt | HOTAIR::MADDOX | When in doubt, change it | Thu May 21 1992 17:13 | 35 |
| This really is an attempt to be helpful, not argumentative.
Yes, you can limit user's access to another system by removing NETMBX
but this won't have the desired results as I understand them. Those
are:
1) Allow users to access some files on the other system but not others
2) Allow users to write to some directories on the other system but not
others.
As I see it there are two obvious ways to accomplish this:
1) Maintain a directory of which files a user can read and into which
directories he can write. Make all users captive ALL-IN-1 accounts.
Customize the heck out of ALL-IN-1 to check every place where a
user can enter a filename to see if the attempted access is allowed.
This will take several man-months to implement and not be fool-proof.
2) Clean up the other system.
You said this couldn't happen by the deadline but have you considered
all the options? Wouldn't it be possible to globaly place an ACL on
all files on the other system which disallowed access to the default
DECNET account, and then go back and allow read and/or write access
to specific directories and files. It seems to me that this might not
be that difficult (I've haven't actually tried it).
I don't mean to stifle imaginative conversation and I'm sure I haven't
thought of everything but it seems to me that you're best long-term
AND short-term solution will have to revolve around fixing security on
the other machine.
Good luck,
Joe
|
231.14 | Thanks for the helpful discussion | THEBAY::SWANDBY | | Thu May 21 1992 17:38 | 17 |
| Thanks for all the input.
As I've dug more into this, it's occurred to me, too, that fixing the
other system would not only be the right thing, but the speedier thing,
too. Guess I'll have to see if politics allow it.
In the meantime, I'll be working up an analysis for the customer
pointing out all the holes we have to plug and how much time it will
take to plug them. Since all this work on our side should be a
temporary fix till the other system gets it together, they may just
decide to use the sledge hammer approach for the short term. It
doesn't make much sense to invest person-months on something we need
because of artificial, temporary constraints.
The discussion has been helpful. Thanks again.
Joel
|