Re: Symantec AntiVirus and AOL

["Joshua Levitsky" <jlevitsk () joshie com>]

Some more research I've done on the AOL adapter has shown me the below. It would seem that Symantec should only be 
using adapters that use the NCF_PHYSICAL flag to select an adapter for GUID generation of just looking at the binding 
order. I'm no driver wizard so if anyone has experience with device drivers then am I right that once a vendors 
intermediate device driver sets the NCD_VIRTUAL bit then they have done their duty as far as saying they are a fake 
adapter? Is the burden then on Symantec to check the flags of adapters they are selecting? I have very little 
experience with network driver design and Windows requirements for driver registration, but the below makes me think 
that the burden is on Symantec to correct something rather than AOL to modify their WAN adapter. Any thoughts would be 
appreciated.

> From wanatw.inf  which is the AOL adapter I found the below. This file can be found at C:\Program Files\Common 
> Files\aol\ACS if you have AOL installed;

[WANATW.ndi.NT]
AddReg = wanatw.w2k.reg
CopyFiles = wanatw.w2k.CopyFiles
Characteristics = 0x09 ; NCF_VIRTUAL | NCF_HIDDEN

[WANATW.ndi.NT.Services]
AddService = wanatw, 2, wanatw.Service, wanatw.EventLog

[wanatw.w2k.reg]
HKR, Ndi, Service, 0,"WANATW"
HKR, Ndi\Interfaces, UpperRange, 0, "ndiswan"
HKR, Ndi\Interfaces, LowerRange, 0, "nolower"
HKR, , WanEndPoints,        0x00010001, "1000"

-----------------------------

> From Micorosft's documentation on what NCF_VIRTUAL and NCF_HIDDEN are;

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/network/hh/network/501install_4q5j.asp

      0x1 NCF_VIRTUAL Component is a virtual adapter. 

      0x8 NCF_HIDDEN Component should not be shown in any user interface. 



  ----- Original Message ----- 
  From: Joshua Levitsky 
  To: full-disclosure () lists netsys com ; Windows NTBugtraq Mailing List ; focus-ms () securityfocus com 
  Sent: Saturday, October 25, 2003 1:09 AM
  Subject: [Full-disclosure] Symantec AntiVirus and AOL


  NOTE: This is not an exploit. This is not a vulnerability. This is simply a bug that makes management of clients more 
difficult / broken. Posting this to hopefully bring a bug in to the open that may not have been discovered yet at 
companies other than my own. If this was a flaw that allowed an exploit or if this allowed a system compromise then I 
would not post this. If anyone knows of other virtual adapters that cause this problem then I would appreciate emails 
listing the product and the MAC address that the product uses.

  ---------------------------

  I have a question for all of you about something I noticed with Symantec AntiVirus Corporate Edition. The SAV CE 
client uses a GUID (unique identifier) that is generated based on the MAC address of the Ethernet adapter on the 
machine SAV CE is running on. This GUID is then used by the SAV CE Parent Server that the managed client checks in to. 
If more than one machine has the same MAC address then in the SSC (Symantec System Console) you will only see 1 machine 
even if 50 machines have the same GUID and check in to the Parent. If you think that your SSC is showing less clients 
than it should then you should possibly read this document to see if this affects you. What I noticed was that this 
problem was not in my corporate image at first but then it happened, and finally enough machines have been replaced / 
reimaged that I noticed I was short on clients in the SSC.

  What is a GUID? Where is the GUID found? A GUID is a Global Unique ID. Well at least it is supposed to be unique. It 
is found in the key below. It is generated using a Microsoft library that is part of RPC. It is based on the MAC 
address of an Ethernet adapter and your current IP address. Is the IP address enough to make it unique even when the 
MAC is the same? I do not know. I have been unable to find the answer. If the IP is enough to make this value unique 
then remediation of this issue is as simple as deleting the GUID key on all workstations that are not showing up in the 
SSC.

  [HKEY_LOCAL_MACHINE\SOFTWARE\Intel\LANDesk\VirusProtect6\CurrentVersion]
  "GUID"=hex:3b,13,fe,fd,4c,6f,cc,4c,89,5b,82,1f,2b,c4,a0,43

  How does SAV CE pick which adapter to make a GUID from? It picks the first thing that looks like an Ethernet adapter 
by going through the binding order in windows. 

  How can multiple machines end up with the same GUID? Imagine a corporate standard image that might include the AOL 
client which adds a hidden adapter to the system. Now when you ghost that image to a new machine and the machine goes 
through sysprep it will add the Ethernet card of the machine you are putting your image on. Since the AOL adapter never 
goes away then the AOL adapter of course comes before the real Ethernet adapter. This means that if the AOL adapter was 
selected by SAV CE in the master image then no new GUID will generate because the GUID only changes when an adapter is 
added or removed. Since the AOL adapter never leaves then it has a good chance of coming before real adapters in the 
binding order. If the AOL adapter was not selected in the master image by the LocalMAC registry key then a GUID should 
be generated when you image a new target machine because SAV CE sees that the old adapter leaves the machine and then 
it will look to the binding order to pick a new adapter. SAV CE hopes to get a real Ethernet card and a GUID should be 
generated based on the MAC address. So this problem happens most when the AOL Adapter is watched by SAV, and that never 
leaves so LocalMAC will always watch it and a new GUID never is generated. 

  How can you see machines with duplicate GUIDs if you think they are not showing up in your SSC? Delete the key below 
using whatever you have like SMS / NetOctopus / etc. The GUID will regenerate and maybe you won't have duplicate GUIDs, 
but if you are generating the GUID from the AOL MAC address then the potential for duplicates continues to exist. Note 
that you want to probably clear out all the host records in the SSC if you do this because you will see duplicates 
because the GUID is how the SSC tracks records, and it takes 30 days for a stale record to purge out of the SSC with 
the current version of SAV CE.

  [HKEY_LOCAL_MACHINE\SOFTWARE\Intel\LANDesk\VirusProtect6\CurrentVersion]
  "GUID"=hex:3b,13,fe,fd,4c,6f,cc,4c,89,5b,82,1f,2b,c4,a0,43

  How can you know that you are using the AOL MAC address instead of the real one? It would seem that AOL 7.x, 8.x, and 
9.x all use 00-03-8a-00-00-15 as the MAC address for the virtual adapter. This aparently is the pseudo-VPN adapter used 
when you connect to AOL over IP. Deleting this key and rebooting will simply result in the same MAC address being 
tracked because the AOL adapter will still come first in the binding list. The binding list is from Windows binding 
order, and not anything from Symantec.

  [HKEY_LOCAL_MACHINE\SOFTWARE\Intel\LANDesk\VirusProtect6\CurrentVersion]
  "LocalMAC"=hex:00,03,8a,00,00,15

  Can't I just change the binding order in Properties for My Network Places in the Advanced Options? No. The AOL 
adapter does not appear there for AOL 7.x, 8.x and 9.x. The adapter does not exist like it did in AOL 6.x. The only way 
to remove the adapter is to uninstall AOL, and even then if you re-install AOL it appears that it can come first again 
in the binding order. You will get a new GUID, but since it is based on the AOL MAC addres then it is possibly you will 
get a duplicate GUID again.

  Doesn't the Parent Server resolve GUID conflicts and tell a client to make a new one if there is a conflict? No. 
Because the GUID is based on the MAC address, and the MAC address should never be the same from machine to machine, the 
Parent Server does not do conflict resolution of GUIDs. I am filing this idea as a RFE however with Symantec because if 
the server simply did this then it would fix issues with other adapters.

  Is this problem AOL's because of the adapter? No. AOL is creating a virtual adapter just like a VPN client might. In 
concept what AOL is doing is fine. The potential for this problem with other pre-loaded software on a master image is 
great if that software makes a virtual adapter. In fact if your company pre-loads VPN software or such then I would 
encourage you to look at the LocalMAC value listed above and see if it matches your real Ethernet adapter that you can 
see by going to Start -> Run and typing "cmd /k ipconfig /all"

  What versions of SAV CE does this affect, and what operating systems? It affects everything from NAV 7.61 through SAV 
8.1.1 that I've looked at, and has the same problem on Windows 2000 and XP. It appears to be a universal problem. 
(Windows NT and 98 were not looked at because they are dead products.)

  Is this problem Symantec's because they bind to a VPN adapter? Sort of but not really. Symantec could have tested for 
more things like default route or adapter able to reach the Parent server when doing the figuring out, but they seem to 
have a pretty good detection system. It would seem to me that the burden is on Symantec to put in an exclusion for the 
AOL MAC address just like any other adapter found to cause similar problems.

  How critical is this? It appears to be more of an annoyance than anything else. If you turn on debugging on your 
clients then you will see the clients check in with "mommy" and it seems like policies are being communicated from the 
Parent servers to their children. So it makes it so that you can't really tell if your environment is protected. 
Depending on how you view this to be critical will make it more or less important to you. 

  What can you do to fix this? I opened a ticket with Symantec Platinum support a week ago. They said they have not 
seen other customers with this issue. They are working on it though and understand the issue, but I think they might 
not be 100% on board with this being something for Symantec to fix. If you find that this issue affects you then I 
encourage you to open a ticket with Symantec so that they can find out if this affects more people than just my 10,000 
clients. While 10,000 is a nice big number, it is only one company so it is understandable that Symantec would wonder 
why no other incidents were filed. If you have this problem then it is urgent that you make a ticket. If you do not 
have support and it would cost you money to make a ticket then please just email me how many clients you have and what 
company you are with, and I will tell Symantec so you do not have to pay for support on this issue.

  Should you be angry with Symantec about this? No. Seriously. There's no way they could have seen this bug. I just 
wanted to post this document out there so that perhaps more feedback could come in to the Symantec support system so 
that perhaps the issue would be given the proper treatment when it gets to the programming staff as a bugfix. 

  Should you be angry with AOL about this? No. Seriously. They make a VPNish adapter to tunnel you in to the network. 
They do everything in a pretty legitimate way, and because they use the same MAC address from version to version it is 
pretty simple to code an exclusion for that MAC address. One could argue that AOL should be using unique MAC addresses, 
but imagine all the MAC addresses that would be burned up by every single install of AOL on every OS using up a MAC 
address. It would be very wasteful.


  --
  Joshua Levitsky, MCSE, CISSP
  System Engineer
  Time Inc. Information Technology
  [5957 F27C 9C71 E9A7 274A 0447 C9B9 75A4 9B41 D4D1]