Agent-based FSSO

Introduction to agent-based FSSO

Fortinet Single Sign-On (FSSO), through agents installed on the network, monitors user logons and passes that information to the FortiGate unit. When a user logs on at a workstation in a monitored domain, FSSO:

l detects the logon event and records the workstation name, domain, and user, l resolves the workstation name to an IP address, l determines which user groups the user belongs to, l sends the user logon information, including IP address and groups list, to the FortiGate unit, l and creates one or more log entries on the FortiGate unit for this logon event as appropriate.

When the user tries to access network resources, the FortiGate unit selects the appropriate security policy for the destination. If the user belongs to one of the permitted user groups associated with that policy, the connection is allowed. Otherwise the connection is denied.

FSSO can also provide NTLM authentication service for requests coming from FortiGate. SSO is very convenient for users, but may not be supported across all platforms. NTLM is not as convenient, but it enjoys wider support.

Introduction to FSSO agents

There are several different FSSO agents that can be used in an FSSO implementation:

l Domain Controller (DC) agent l eDirectory agent l Citrix/Terminal Server (TS) agent l Collector agent

Consult the latest FortiOS and FSSO Release Notes for operating system compatibility information.

Domain Controller (DC) agent

The Domain Controller (DC) agent must be installed on every domain controller if you will use DC Agent mode, but is not required if you use Polling mode.

eDirectory agent

The eDirectory agent is installed on a Novell network to monitor user logons and send the required information to the FortiGate unit. It functions much like the Collector agent on a Windows AD domain controller. The agent can obtain information from the Novell eDirectory using either the Novell API or LDAP.

Terminal Server (TS) agent

Terminal Server (TS) agent can be installed on a Citrix or VMware Horizon 7.4 terminal server to monitor user logons in real time. It functions much like the DC Agent on a Windows AD domain controller.

Collector agent

This agent is installed as a service on a server in the Windows AD network to monitor user logons and send the required information to the FortiGate unit. The Collector agent can collect information from

l Domain Controller agent (Windows AD) l TS agent (Citrix or VMware Horizon Terminal Server)

In a Windows AD network, the Collector agent can optionally obtain logon information by polling the AD domain controllers. In this case, DC agents are not needed.

The Collector can obtain user group information from the DC agent or optionally, a FortiGate unit can obtain group information directly from AD using Lightweight Directory Access Protocol (LDAP).

On a Windows AD network, the FSSO software can also serve NT LAN Manager (NTLM) requests coming from client browsers (forwarded by the FortiGate unit) with only one or more Collector agents installed.

The CA is responsible for DNS lookups, group verification, workstation checks, and as mentioned FortiGate updates of logon records. The FSSO Collector Agent sends Domain Local Security Group and Global Security Group information to FortiGate units. The CA communicates with the FortiGate over TCP port 8000 and it listens on UDP port 8002 for updates from the DC agents.

The FortiGate unit can have up to five CAs configured for redundancy. If the first on the list is unreachable, the next is attempted, and so on down the list until one is contacted.

All DC agents must point to the correct Collector agent port number and IP address on domains with multiple DCs.

A FortiAuthenticator unit can act much like a Collector agent, collecting Windows AD user logon information and sending it to the FortiGate unit. It is particularly useful in large installations with several FortiGate units. For more information, see the FortiAuthenticator Administration Guide.

FSSO for Microsoft Exchange Server

As of FortiOS 5.4, FSSO supports monitoring Microsoft Exchange Server. This is useful for situations when the user accesses the domain account to view their email, even when the client device might not be in the domain.

Support for the Exchange server is configured on the Back-end FSSO collector agent. For more information on the collector agent, see Collector agent installation:

1. On the FSSO collector agent, go to Advanced Settings > Exchange Server.
2. Select Add and enter the following information and select OK:

Domain Name	Enter your domain name.
Server IP/Hostname	Enter the IP address or the hostname of your exchange server.
Polling forwarded event log	This option for scenarios when you do not want that CA polls the Exchange Server logs directly. In this case you need to configure event log forwarding on the Exchange server. Exchange event logs can be forwarded to any member server. If you enable this, instead of the IP of the Exchange server configured in the previous step, you must then configure the IP of this member server. CA will then contact the member server.
Ignore Name	Because CA will also check Windows log files for logon events and when a user authenticates to Exchange Server there is also a logon event in Windows event log, which CA will read and this will overwrite the Exchange Server logon event (ESEventLog) on CA. So it is recommended to set the ignore list to the domain the user belongs to. To do so, enter the domain name in the Ignore Name field and select Add.

FSSO for Windows AD

FSSO for Windows AD requires at least one Collector agent. Domain Controller agents may also be required depending on the Collector agent working mode. There are two working modes to monitor user logon activity: DC Agent mode or Polling mode.

Collector agent DC Agent mode versus Polling mode

	DC Agent mode	Polling Mode
Installation	Complex — Multiple installations: one agent per DC plus Collector agent, requires a reboot	Easy — Only Collector agent installation, no reboot required
Resources	Shares resources with DC system	Has own resources
Network load	Each DC agent requires minimum 64kpbs bandwidth, adding to network load	Increase polling period during busy period to reduce network load
Level of Confidence	Captures all logons	Potential to miss a login if polling period is too great

DC Agent mode

DC Agent mode is the standard mode for FSSO. In DC Agent mode, a Fortinet authentication agent is installed on each domain controller. These DC agents monitor user logon events and pass the information to the Collector agent, which stores the information and sends it to the FortiGate unit.

The DC agent installed on the domain controllers is not a service like the Collector agent — it is a DLL file called dcagent.dll and is installed in the Windows\system32 directory. It must be installed on all domain controllers of the domains that are being monitored.

FSSO in DC agent mode

DC Agent mode provides reliable user logon information, however you must install a DC agent on every domain controller. A reboot is needed after the agent is installed. Each installation requires some maintenance as well. For these reasons it may not be possible to use the DC Agent mode.

Each domain controller connection needs a minimum guaranteed 64kpbs bandwidth to ensure proper FSSO functionality. You can optionally configure traffic shapers on the FortiGate unit to ensure this minimum bandwidth is guaranteed for the domain controller connections.

Introduction to agent-based

Polling mode

In Polling mode there are three options — NetAPI polling, Event log polling, and Event log using WMI. All share the advantages of being transparent and agentless.

NetAPI polling is used to retrieve server logon sessions. This includes the logon event information for the Controller agent. NetAPI runs faster than Event log polling but it may miss some user logon events under heavy system load. It requires a query round trip time of less than 10 seconds.

Event log polling may run a bit slower, but will not miss events, even when the installation site has many users that require authentication. It does not have the 10 second limit on NetAPI polling. Event log polling requires fast network links. Event log polling is required if there are Mac OS users logging into Windows AD.

Event log using WMI polling: WMI is a Windows API to get system information from a Windows server, CA is a WMI client and sends WMI queries for user logon events to DC, which in this case is a WMI server. Main advantage in this mode is that CA does not need to search security event logs on DC for user logon events, instead, DC returns all requested logon events via WMI. This also reduces network load between CA and DC.

In Polling mode, the Collector agent polls port 445 of each domain controller for user logon information every few seconds and forwards it to the FortiGate unit. There are no DC Agents installed, so the Collector agent polls the domain controllers directly.

FSSO in polling mode

A major benefit of Polling mode is that no FSSO DC Agents are required. If it is not possible to install FSSO DC Agents on your domain controllers, this is the alternate configuration available to you. Polling mode results in a less complex install, and reduces ongoing maintenance. The minimum permissions required in Polling mode are to read the event log or call NetAPI.

Collector agent AD access mode – standard versus advanced

The Collector agent has two ways to access Active Directory user information. The main difference between standard and advanced mode is the naming convention used when referring to username information.

Standard mode uses regular Windows convention: Domain\Username. Advanced mode uses LDAP convention: CN=User, OU=Name, DC=Domain.

If there is no special requirement to use LDAP— best practices suggest you set up FSSO in standard mode. This mode is easier to set up, and is usually easier to maintain and troubleshoot.

Standard and advanced modes have the same level of functionality with the following exceptions:

• Users have to create Group filters on the Collector agent. This differs from advanced mode where Group filters are configured from the FortiGate unit. Fortinet strongly encourages users to create filters from CA.
• Advanced mode supports nested or inherited groups. This means that users may be a member of multiple monitored groups. Standard mode does not support nested groups so a user must be a direct member of the group being monitored.

FSSO for Citrix

Citrix users can enjoy a similar Single Sign-On experience as Windows AD users. The FSSO TS agent installed on each Citrix server provides user logon information to the FSSO Collector agent on the network. The FortiGate unit uses this information to authenticate the user in security policies.

Citrix SSO topology

Citrix users do not have unique IP addresses. When a Citrix user logs on, the TS agent assigns that user a range of ports. By default each user has a range of 200 ports.

FSSO for Novell eDirectory

FSSO in a Novell eDirectory environment works similar to the FSSO Polling mode in the Windows AD environment. The eDirectory agent polls the eDirectory servers for user logon information and forwards the information to the FortiGate unit. There is no need for the Collector agent.

When a user logs on at a workstation, FSSO:

 

FSSO NTLM authentication support

l detects the logon event by polling the eDirectory server and records the IP address and user ID, l looks up in the eDirectory which groups this user belongs to, l sends the IP address and user groups information to the FortiGate unit.

When the user tries to access network resources, the FortiGate unit selects the appropriate security policy for the destination. If the user belongs to one of the permitted user groups, the connection is allowed.

FSSO is supported on the Novell E-Directory 8.8 operating system.

For a Novell network, there is only one FSSO component to install — the eDirectory agent. In some cases, you also need to install the Novell Client.

FSSO security issues

When the different components of FSSO are communicating there are some inherent security features.

FSSO installation requires an account with network admin privileges. The security inherent in these types of accounts helps ensure access to FSSO configurations is not tampered with.

User passwords are never sent between FSSO components. The information that is sent is information to identify a user including the username, group or groups, and IP address.

NTLM uses base-64 encoded packets, and uses a unique randomly generated challenge nonce to avoid sending user information and password between the client and the server.

FSSO NTLM authentication support

In a Windows AD network, FSSO can also provide NTLM authentication service to the FortiGate unit. When the user makes a request that requires authentication, the FortiGate unit initiates NTLM negotiation with the client browser. The FortiGate unit does not process the NTLM packets itself. Instead, it forwards all the NTLM packets to the FSSO service to process.

NTLM has the benefit of not requiring an FSSO agent, but it is not transparent to users, and the user’s web browser must support NTLM.

The NTLM protocol protects the user’s password by not sending it over the network. Instead, the server sends the client a random number that the client must encrypt with the hash value of the user’s password. The server compares the result of the client’s encryption with the result of its own encryption. The two will match only if both parties used the same password.

FSSO NTLM authentication support

NTLM authentication

If the NTLM authentication with the Windows AD network is successful, and the user belongs to one of the groups permitted in the applicable security policy, the FortiGate unit allows the connection but will require authentication again in the future when the current authentication expires.

Fortinet has tested NTLM authentication with Internet Explorer and Firefox browsers.

NTLM in a multiple domain environment

In a multiple domain environment for NTLM, the important factor is that there is a trust relation between the domains. In a forest, this relation is automatically created. So you can install FSSO agent on one of the domain controllers without worry.

But in case of multiple domains that are not in a forest, you need to create a trust relation between the domains. If you do not want to have a trust relation between your multiple domains, you need to use FSAE 4.0 MR1 and the DC agent needs to be installed once on each domain. Then you can use security policies to configure server access.

In the figure below, three domains are shown connected to the FSSO Collector agent server. The Client logs on to their local Domain Controller, which then sends the user logon event information to the Collector Agent. When the Client attempts to access the Internet, the FortiGate unit contacts the Collector Agent for the logon

FSSO NTLM authentication support

information, sees the Client is authenticated, and allows access to the Internet. There are multiple domains each with a domain controller agent (DCagent) that sends logon information to the Collector agent. If the multiple domains have a trust relationship, only one DCagent is required instead of one per domain.

FSSO NTLM with multiple domains not in a forest

Understanding the NTLM authentication process

1. The user attempts to connect to an external (internet) HTTP resource. The client application (browser) on the user’s computer issues an unauthenticated request through the FortiGate unit.
2. The FortiGate is aware that this client has not authenticated previously, so responds with a 401

Unauthenticated status code, and tells the client which authentication method to reply with in the header: Proxy-Authenticated: NTLM. Then the initial session is dismantled.

3. The client application connects again to the FortiGate, and issues a GET-request, with a

Proxy-Authorization: NTLM <negotiate string> header. <negotiate-string> is a base64encoded NTLM Type 1 negotiation packet.

4. The FortiGate unit replies with a 401 “proxy auth required” status code, and a

Proxy-Authenticate: NTLM <challenge string> (a base 64-encoded NTLM Type 2 challenge packet). In this packet is the challenge nonce, a random number chosen for this negotiation that is used once and prevents replay attacks.

The TCP connection must be kept alive, as all subsequent authentication-related information is tied to the TCP connection. If it is dropped, the authentication process must start again from the beginning.

5. The client sends a new GET-request with a header: Proxy-Authenticate: NTLM <authenticate string>, where <authenticate string> is a NTLM Type 3 Authentication packet that contains:

 

l username and domain l the challenge nonce encoded with the client password (it may contain the challenge nonce twice using different algorithms).

6. If the negotiation is successful and the user belongs to one of the groups permitted in the security policy, the connection is allowed, Otherwise, the FortiGate unit denies the authentication by issuing a 401 return code and prompts for a username and password. Unless the TCP connection is broken, no further credentials are sent from the client to the proxy.