cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
4983
Views
4
Helpful
3
Replies

ACS 5.2 and AD Integration

lambchopper
Level 1
Level 1

I have a few questions about AD integration of the ACS 5.2 server:

The ACS server documentation states the following:

"You must enter the username and password credentials in the ACS 5.2 configuration for ACS to join and communicate with the AD domain. The credentials must have sufficient permissions to create a computer object. If a user's AD group membership and attribute information is required for access policy, they must first be selected in the AD configuration."

Obvously our AD Administrators are a little concerned about giving write access to this account.  Am I correct in thinking that the need for "sufficient permissions to create a computer object" are only used by the ACS at the time it joins the domain?

Can the create computer object permissions be revoked after the ACS has joined the domain to tighten security?

What other than creating this computer object would the ACS account need write access for within AD if this permission can not be revoked?

I'm not sure I understand the meaning of the the sentence: "If a user's AD group membership and attribute information is required for access policy, they must first be selected in the AD configuration."  Could someone provide some clarity?

The ACS Server documentation says:

"ACS 5.2 joins Active Directory (AD) directly and does not rely on a domain-joined Windows Server. ACS Remote Agent is not required."

Can I assume this means the ACS server joins the domain as any Windows client would be joined (E.G. a standard computer object) could someone confirm this?

How does the ACS server (Linux based) read AD objects?  Keberos or ADSI?

Thank you in advance!

3 Replies 3

robdowson
Level 1
Level 1

Hi Dave,

We had the same questions when we were adding our 5.2 to the domain aswell.

It seems like it's just the same as adding a windows PC to the domain - ie. the administrator account needs access to define the new computer object - but it doesn't look like the administrator account is used after that - although it does continue to show up in the web-interface - but I've tried eg. disabling the admin account we used - and ACS still continued to work.

Rather than use a normal admin account though just in case / to stop it looking odd in the web-interface, our AD administrators created a service account specifically for ACS to use to join the domain.

They created a new OU to contain the new computer objects - eg. Servers\ACS Appliances\

...and gave the service account delegated right to the computer objects in the new OU.

So since it only has access to that OU - I don't believe there's an issue with leaving the account enabled - but like I said - I don't think it's used apart from during the 'join' process - so you could disable it. I've never had to re-join the domain since the original join.

We did have some problems joining the domain though - despite that setup. I think our AD guys had to manually create the computer objects first (normally automatically created by the ACS joining) - and then "delegate control for the ACS service user to join computers in the target OU by creating a custom task to delegate" - I believe they used the 'Delegation of control wizard' in AD users and computers, selected the role account, 'create a custom task to delegate', 'only the following objects in the folder' - computer objects - create selected objects in this folder & delete selected objects in this folder, then Permissions - tick general and select 'validated write to DNS host name' / 'validated write to service principal name' / 'change password' / 'reset password' / 'read and write DNS host name attributes'.

>I'm not sure I understand the meaning of the the sentence: "If a user's AD group membership and attribute information is required for access policy, they must first be selected in the AD configuration."  Could someone provide some clarity?

This isn't very clear is it - but I think this is trying to say that you need to select the things in AD you want to use before you can use them in the policy. ie. we're using group memebership (ie. a user has to be in a certain group in AD to get the correct access) - but you can't just refer to those groups from the policy - you need to select those groups from AD first - ie.

Users and Identity Stores > ... > External Identity Stores > Active Directory - Directory Groups tab - Select.

Then you can use them in the group mapping policy to map to an internal ACS group - which can then be used in the Authorization policy.

>"ACS 5.2 joins Active Directory (AD) directly and does not rely on a domain-joined Windows Server. ACS Remote Agent is not required."

Yes - I beleive ACS prior to 5.x required an agent installing on the AD DC - but now that is not the case as ACS joins AD natively. From some debugging we've done - it appears to use the 'CentrifyDC' linux packages - so it talks to AD natively as if it were a normal AD client.

AD integration is good - when you get it working! Look out for problems with ACS not obeying 'AD sites and services' policy - ie. it won't use the DC you want it to if you've defined AD S&S - it will connect to any DC in the domain - which is interesting if some of your DCs are behind firewalls like ours. Apparently this isn't supported - but there's a feature request in to add it - hopefully soon!

Hope that helps! Good luck!

iilyinas
Level 3
Level 3

Hi, Dave!

Can the create computer object permissions be revoked after the ACS has joined the domain to tighten security?

What  other than creating this computer object would the ACS account need  write access for within AD if this permission can not be revoked?

I'd say, it is possible, but you can easily run into a trouble.

You'll need permissions to be added back in few scenarios:

1. If you'll add new secondaries;

2. If you'll do an upgrade;

3. if you'll do HW replacement with restoration from a backup.

Also, ACS re-discovers AD from time to time. So I'd not recommend to remove rights/object permissions. When using AD identity store, for example, ACS tries to resolve servers (KDC, DC, GC) every 8 hours, using DNS. Beware of the fact, that until DNS resolution is not finished, ACS will not be able to authenticate users using AD. So it's important to have fast response from NS, especially in case of large deployment.


ACS needs permissions to change password (as by default, machine passwords expires after 30 days in AD), and to delete the account (if the AD configuration is removed from the ACS).

I'm  not sure I understand the meaning of the the sentence: "If a user's AD  group membership and attribute information is required for access  policy, they must first be selected in the AD configuration."  Could  someone provide some clarity?


If you want to set access to the user, basing on AD group membership, you have to select them in AD configuration, as Rob wrote. After that, ACS will fetch them from AD and check against them, when needed.

The ACS Server documentation says:

"ACS  5.2 joins Active Directory (AD) directly and does not rely on a  domain-joined Windows Server. ACS Remote Agent is not required."

Can  I assume this means the ACS server joins the domain as any Windows  client would be joined (E.G. a standard computer object) could someone  confirm this?

Yes, it joins the domain as any regular server. This part of the documentation refers to the fact that ACS 4, installed on Windows, uses another mechanism to join to the domain.

How does the ACS server (Linux based) read AD objects?  Keberos or ADSI?

It uses Centrifydc, as correctly mentioned by Rob. This component is not configurable by ACS CLI, and normally not accessible.

Cheers, Irina

--

Please, rate helpful posts!

Naveen Kumar
Level 4
Level 4


Just to Share:

 
ACS 5.x and later: Integration with Microsoft Active Directory Configuration Example:

http://www.cisco.com/c/en/us/support/docs/security/secure-access-control-system/113571-acs5-ad-int-config-00.html