Home > Articles > Operating Systems, Server > Microsoft Servers

  • Print
  • + Share This
This chapter is from the book

Security

There probably hasn't been a time in the modern era of computing where there was more widespread concern for security. Only a few years ago, security was one of those dark corners of the industry for hard-core types, but largely ignored by most of us. Today, securing computing resources ripples from the largest corporation to the most computer-illiterate home consumer, trying to prevent sophisticated hackers not only from infecting their computers with destructive viruses, but also from stealing information or using their computer as an ad hoc server.

NOTE

Jan de Clercq of HP's Security Office provided much of the technical information and technical review of this section. Jan is one of HP's leading security experts. I highly recommend his new book, Windows Server 2003 Security Infrastructures, Digital Press, 2004.

Windows 2000 made some huge strides in securing the computing environment with features such as Kerberos for authentication and authorization, IP Security (IPSec) for remote access security, and the Encrypted File System (EFS) for protecting sensitive data. Windows NT and 2000 provided Certificate Services, which gave Administrators the ability to install their own CA and secure the certificates in their enterprise inexpensively, enticing smaller companies to implement Public Key Infrastructures (PKI).

Security issues always provide room for improvement, and Windows Server 2003 and Windows XP have taken a step forward in a number of areas. Think of XP as a Windows 2003 client—they both have very similar and compatible features. Significant security feature improvements in Windows XP and Windows Server 2003 include

  • Software restriction policies

  • Internet connection firewall

  • Personal firewall

  • EFS enhancements

  • IPSec enhancements

  • Credential Manager

  • PKI improvements

    • Cross-certification trust model

    • Editable certificate templates (v2)

    • Key recovery

  • Enhanced Security Management—Effective Permissions tab

  • User Autoenrollment

  • Kerberos and forest trusts

The following sections provide a brief description of each issue. Refer to Jan de Clercq's books listed in the "References" section at the end of the chapter for more information.

Software Restriction Policies

Also known as SAFER, these policies are Microsoft's first attempt to provide stronger cryptographic compliance with Federal Information Processing Standards (FIPS) 140-1. FIPS determines whether products meet the standard and evaluates overall security provided by a cryptosystem. The kernel mode driver, FIPS.SYS, supports EFS (efs.sys), IPSec (ipsec.sys), and other crypto functions. Although this is only a level 1 compliance intended for general PC use, it is nonetheless a start. The idea with SAFER policies is to provide the Administrator the ability to grant granular client-level security policies to restrict the access to read, write, or execute software on computers.

Because the "How To" basics of configuring these policies are reasonably well documented in the Microsoft KB 324036 article "How to use Software Restriction Policies in Windows Server 2003," I'll use the space here to describe some important concepts and implementation recommendations.

Software Restriction Policies is implemented through GPO settings located in the GPO tree Computer Configuration\Windows Settings\Security Settings\Software Restriction Policies, and in User Configuration\Windows Settings\Security Settings\Software Restriction Policies. When you drill down to this location on a new policy, a warning appears in the right pane of the GPO Editor, "No Software Restriction Policies Defined," as shown in Figure 1.23. Right-click on the Software Restriction Policies in the left pane and click New Software Restriction Policies. You will then see two folders titled Security Levels and Additional Rules. Security Levels defines a high-level access policy, whereas Additional Rules allows you to define specific rules, including the following:

  • Path rules: Allow or deny access to files specified by a directory path on the local PC. Also available is a Registry path. Using the Registry path to identify files to allow or deny access to is more reliable. For instance, if you deny access to C:/applications/abc.exe to prevent users from running the ABC program, but the user installs it on D:\myapps\abc.exe, the policy would fail. The Registry path for any given application will be the same no matter where the files are stored on disk. This can also be used to disallow access to a known file name that is installed by a virus.

  • Hash rules: The Administrator can generate a hash of a file and set up a hash rule to compare the original hash to a hash of the existing file. If they differ, access is denied. This rule is intended to allow users to run or prevent them from running certain versions of a program. This could also be useful in preventing users from accessing files infected with a virus. The hash is not affected by file rename or relocation.

  • Certificate rules: These rules can be used to allow or disallow access to a digitally signed program. A comparison is made via user certificate to determine access level.

  • Internet Zone rule: As of this writing, this rule is pretty useless in my opinion. Its intent is to let you create a rule to prevent users from installing software from sites in a given Internet Zone defined in Internet Explorer (IE). Unfortunately, it applies only to downloading .MSI files—not a big threat since there are so many other file types: .exe, .zip, etc..

Security Levels

Think of Security Levels as the locked condition on the front door to a building. There are two options (look in the Security Levels folder in the GPO): Disallowed and Unrestricted. The Disallowed option locks the front door and requires all who enter to provide proper credentials. No public access. Employees with ID badges only. The Unrestricted option leaves the front door unlocked, allowing anyone to enter the building. Inside the building are special rooms (programs and files on the PC) that also have locks on the door. These locks are the four classes of rules described in the previous section. If the front door is Disallowed, then you don't get access unless you can prove you have access to a room in the building. The guard escorts you to that room (or rooms), but you can't go anywhere else. This obviously is a high-security condition. If the front door is set to Unrestricted, anyone can walk in and roam through the building and enter any room that is not locked. However, if a room is locked (that is, a rule has been created to restrict access to a program), you have to be granted specific access to get in. I recommend, as does Microsoft's KB 310791 article "Description of the Software Restriction Policies in Windows XP," you use the Unrestricted option because it simply locks out certain files and programs. Unrestricted is the default. Setting the Disallowed mode, unless it is in a very controlled environment, will generate a lot of help desk calls if it's not well planned and tested.

Figure 1.23Figure 1.23 Defining Software Restriction Policies in the Group Policy Editor.

Additional Rules

These are the keys to the rooms—the exceptions to the lockout. Note that four Registry path policies are created by default in the Additional Rules folder. Figure 1.24 shows the default Registry path rules, as viewed in the GPMC. They are all defined as Unrestricted and are applied to

%systemroot%
%systemroot%\*.exe
%systemroot%\system32\*.exe
%programfilesdir%\

The purpose of these policies is to permit access to critical directory space when you've set the Security Level to Disallowed. These policies allow access to areas like c:\> Program files.

Figure 1.24Figure 1.24 The additional rules defined by default in Software Restriction Policies are shown using the GPMC snap-in.

Implementation

So far these rules seem fairly straightforward. If you want to put the admin pack on every PC, but restrict users from running it, you set the SAFER policy to Unrestricted, and then create a path policy to restrict domain users from executing it. Pretty simple until you get multiple rules defined in multiple policies and perhaps in a combination of User and Computer Configuration settings.

NOTE

The description of functionality described here is based on my experience and testing and asking questions of Microsoft PSS. I have not seen any Microsoft documentation to confirm or deny these assertions. Your mileage may vary, so be sure you test these policies before implementing.

If multiple Software Restriction Policies are defined, things get complicated. Several design rules apply to the application of these policies. Understanding them is critical to obtaining the desired results.

  • Closest match applies the rule. That is, if Rule 1 allows access to c:\*.exe and Rule 2 denies access to C:\payrollapp.exe, then the user would be unable to run payrollapp.exe because Rule 2 is a closer match to the program being run.

  • Most restrictive rules apply. Apply most-restrictive path rules further down in the directory tree than less-restrictive ones.

  • Rules are processed in the following order:

    1. Security Level (highest priority)

    2. Hash

    3. Certificate

    4. Path

    5. Internet Zone rule

    6. Default rules

  • Software Restriction Policies accumulate from multiple GPOs and are evaluated together. They do not obey normal SDOU (Site Domain Organizational Unit) processing of GPOs, and they can be applied at the Computer and User Configuration sections of each GPO.

  • Security Levels determine how the Additional Rules are defined. If the Security Level is Unrestricted, the Additional Rules should be designed to restrict access. If the Security Level is Disallowed, the Additional Rules should be designed to allow access.

  • Software restriction policies—especially when setting general categories (such as entire directories) to be disallowed—can have unexpected results. For instance, disallowing access to *.vbs to prevent virus deposited files from running can also restrict access to logon scripts and other valid scripts.

  • Always thoroughly test these policies.

  • Define the policies in individual GPOs. This makes them easier to manage and troubleshoot.

  • Avoid applying Software Restriction Policies from multiple locations (domains and OUs) if possible. Applying them from fewer locations makes the results easier to predict and troubleshoot.

  • To restrict Administrators from a Software Restriction Policy defined in User Configuration, restrict the appropriate Administrators' accounts and/or groups by removing the Apply Group Policy directly on the GPO.

  • To restrict Administrators from a Software Restriction Policy defined in Computer Configuration, you can remove the Apply Group Policy directly from the appropriate user or group, or you can go to Computer Configuration\Windows Settings\Security Settings\Software Restriction Policies\Enforcement. In the Enforcement Properties page, under the Apply Software Restriction policies to the following users, select All Users Except Local Administrators.

  • Make certain you have not inadvertently restricted access to valid files such as logon scripts and directories and files that should be scanned by antivirus programs. Testing should expose these types of flaws.

Evaluation of these policies at this time seems like a bit of black magic. As of this writing, Microsoft had not seen a lot of customers using these policies, but in the work I've done with the policies, I've found they are difficult to design and it's difficult to get the desired results if you have a lot of them. Again, adding complexity to these policies results in a high probability of failure, so you must test them thoroughly. You can apply them to User Config and Computer Config on multiple levels (domain and OU), and then have them all sorted by rule precedence, closest match, and most restrictive. Keep these GPOs in as few domains and OUs as possible, keep the rules simple and well planned, and avoid applying them to both User Config and Computer Config in the GPO.

Sample Rules

Let's look at a couple of examples of rules. Suppose we have six policies defined in a Domain SAFER Policy defined at the domain:

  • Security Level: Unrestricted

  • Hash Rule 1: hash of Payroll.exe (v4.2) set to Unrestricted

  • Hash Rule 2: hash of Payroll.exe (v4.1) set to Disallowed

  • Certificate Rule 1: Applications Certificate set to Unrestricted

  • Path Rule 1: C:\RestrictedApps set to Disallowed

  • Path Rule 2: C:\RestrictedApps\*.exe set to Unrestricted

Using these rules, a user attempting to run the following programs would have these results:

  • User runs the program: C:\restrictedApps\Payroll.exe (v4.1).

  • Security Level set to Unrestricted allows access.

    Path Rule 1 allows access.

    Hash Rule 1 disallows access.

    Certificate Rule 1 applies because this app was digitally signed by the certificate owned by the Applications group and is set to Unrestricted.

    Result: Hash Rule 2 applies because it is most specific (denies access to the exact file name, whereas the others refer to directories or groups of files). Access is granted.

  • User runs the program: C:\restrictedApps\backupfiles.exe.

  • Security Level set to Unrestricted allows access.

    Path Rule 1 disallows access to C:\restrictedApps.

    Path Rule 2 allows access to C:\restrictedApps\*.exe.

    Result: Because Path Rule 1 denies access to the directory, Path Rule 2 never applies. Access is denied. Remember that the most-restrictive path rules should be applied deeper in the tree than the least-restrictive ones.

Suppose the same rules are defined, but the Security Level is set to Disallowed. Path Rule 1 and Hash Rule 2 are not needed because the Security Level is set to Disallowed. With that setting, all you need to do is define rules to allow access to programs that you want the user to access.

Microsoft has attempted to help with some exceptions to these rules. In the Software Restriction Policies folder in the GPO, you'll see three policies besides the two folders. These policies are

  • Enforcement: Allows you to exclude .dlls from restricted programs, and to exclude local admins.

  • Designated File Types: By default, policies only apply to normal executable types, such as .exe, .dll, and .vbs. This option allows you to add other types to it, such as .bat, .cmd, .msi, and others, including a blank so you can add one that isn't in the list.

  • Trusted Publishers: Defines users who can choose trusted publishers (for certificates).

In summary of Software Restriction Policies, I'd say to use them very carefully and only after extensive testing so you understand what the ramifications are. They have some value in the security configuration of the enterprise, but they are complex and can create a lot of problems if you aren't careful.

Some good resources for SAFER policies include the Microsoft whitepaper, and Jeremy Moskowitz's Group Policy, Profiles and IntelliMirror for Windows 2003, Windows XP, and Windows 2000, Sybex, March 2004.

Internet Connection (Personal) Firewall

Also referred to as the Personal Firewall, this feature is provided with Windows XP as built-in protection for each computer. It isn't the most secure firewall, but it's probably sufficient for casual home users. If firewalls are implemented elsewhere, the personal firewall in XP is usually required to not be enabled. To enable the Internet Connection Firewall, open a network connection's Properties page, go to the Advanced tab, and select the Protect My Computer and Network by Limiting or Preventing Access to this Computer from the Internet check box. Click the Settings button to access the Advanced Settings pane, shown in Figure 1.25 to configure advanced settings, including

  • Services: Allows you to select services such as FTP, Telnet, HTTP, and Remote Desktop that the users of this computer are allowed to access. Default is no access. So if you turn this firewall on and can't get to a Web site, the reason is probably wasn't enabled.

  • Security Logging: Allows configuring options to log dropped packets and successful connections (or not), to specify a log file name and location (default is %systemroot%\pfirewall.log), and to set a log file size limit.

  • ICMP: Permits configuring of ICMP traffic, such as Allow Incoming Echo Request, Allow Outgoing Search Request, Allow Incoming Router Request, and other options. By default, all are disabled.

NOTE

Windows XP Service Pack 2 enables the Internet Connection Firewall by default.

EFS (Encrypting File System) Enhancements

Although Windows 2000 EFS brought a level of security to local files, it had a number of deficiencies:

  • Stealing a local account password is easy using common hacker tools in standalone mode.

  • Encrypted files stored on file servers are decrypted on the server and then transported in clear text across the network to the user's workstation. Because EFS needs access to the user's private key, which is held in the profile, the server must be "trusted for delegation" and have access to the user's local profile.

  • Inability to share EFS encrypted files.

Figure 1.25Figure 1.25 Internet Connection Firewall Advanced Settings enable you to define what the user is allowed to do.

EFS has been enhanced in Windows XP and Windows Server 2003 to support sharing of encrypted files. Windows 2000, Windows Server 2003, and Windows XP all store the EFS metadata in the NTFS. Thus, encrypted files on NTFS volumes in Windows 2000 and Windows Server 2003. Sharing an encrypted file is enabled by opening the file's Properties page, clicking the Advanced button, and selecting the Details button in the Advanced Attributes area. You can add users who you want to share the file with to the list.

Note that EFS file sharing is set at the file level—not the folder level——and inheritance of EFS file-sharing metadata is not supported.

Another option for sharing encrypted files in Windows Server 2003 is to use Web folders. Windows Server 2003's support for the transport of EFS metadata using the WebDAV protocol, an extension of HTTP 1.1, makes Web folders an excellent alternative to sharing files on file servers, and is recommended by Microsoft. WebDAV is supported on IE v5.0 on the client and IIS v5.0 and later on the server. For more information about WebDAV, see the WebDAV Resources FAQs at http://www.webdav.org/other/faq.html.

To configure a Web folder on the server running IIS, go to the properties sheet of the encrypted file (or any file), go to the Web Sharing tab, and select the Share this Folder option. Table 1.5 provides a side-by-side comparison of features between the Web folders and File Shares.

Table 1.5 Comparison of EFS Features Between File Shares and Web Folders

Remote EFS Operations On . . .

. . . File Shares

. . . Web Folders

Where does EFS encryption/decryption occur?

Files are encrypted and decrypted on the file server.

Files are encrypted and decrypted on the user machine.

Are the files secured during transmission over the network?

Files are sent in the clear over the network connection.

Files remain encrypted while being sent over the network connection.

What technology is or can be used to secure the transmission of the files over the network?

Requires IPSec to secure the file transfer between file server and user machine.

Does not require IPSec to secure the file transfer; relies on the WebDAV EFS extensions to securely transmit the file.

Must the file server be "trusted for delegation?"

Requires file server to be "trusted for delegation."

Does not require file server to be "trusted for delegation."

Does the solution require a copy of the user profile on the file server?

Requires availability of user profile on the file server (local or roaming profile).

Does not require availability of user profile on the file server.

Where does the EFS file-sharing authorization process take place for users?

EFS checks for other user certificates on the file server and/or in the AD

EFS checks for other user certificates on the local machine and/or in the AD.


(Table reprinted by permission from Jan de Clercq, Hewlett-Packard Company.)

Credential Manager

Single Sign On (SSO) capability has been a frequent request from users and Administrators who find it challenging to keep track of all the credentials needed to log on to the domain, or use Microsoft Passport credentials and Smart Card credentials. Managers who must keep getting validated to perform administrative tasks on various computers also have requested this capability. Windows Server 2003 introduces Credential Manager, a client-based SSO solution that uses an intelligent credential-caching mechanism that keeps credentials in a store on the client, requiring the user to provide a single username/password to open the store. Credentials in the store can consist of a user account and password; a user account, certificate, and private key (which can be stored on a smart card); or Microsoft Passport credentials.

The credential store is part of the user profile and supports roaming and can optionally be disabled by the Administrator via the GPO. This is done by opening a GPO and going to Windows Settings\Security Settings\Local Policies\Security Options and enabling the option Network Access: do not allow storage of credentials or .NET Passports for network authentication.

This feature is available in Windows Server 2003, Windows XP Pro, and Windows XP Home operating systems.

Public Key Infrastructure (PKI) Improvements

Microsoft provided an out-of-the-box, user-defined and administered CA in the Windows NT 4.0 Option pack. This might not have been the most effective and robust product on the market, but it was cheap (and was included at no extra cost with Windows 2000). This allowed small- and medium-size companies, who couldn't afford the high cost of third-party companies such as VeriSign and Entrust, to fortify their security systems. However, according to one of HP's security experts, consultant Jan de Clercq, "ease of use and security don't easily coexist." But considering the improvements made in Windows Server 2003 security features, especially in the PKI arena, Jan stated that "Given the low cost and the advanced PKI features of Windows.NET, it may be a product that will bring PKI to the masses." Let's see what features are now available to make this possible.

Cross Certification Trust Model

Windows 2000 used a hierarchical CA trust model, shown in Figure 1.26, whereby the parent (root) CA was linked to the child (subordinate) CA by trusts. Only one root CA can be in the forest, and the only way to configure it to trust certificates issued by other CAs in other forests was to reconfigure the entire CA structure using Certificate Trust Lists. This might be desirable in multiforest enterprises or to link extranets.

Note that the CA hierarchy is not specifically tied to the domain hierarchy. Rather, it might more appropriately be associated with geographies of the enterprise to provide accessible CAs to users and computers in each geography.

Windows Server 2003 introduces the Cross Certification trust model, which allows CAs in different organizations (forests) to trust each other and is in addition to the subordinate trust model. The Cross Certification model permits the Administrator some freedom in configuring the trust relationships with the extranet partners, by specifying whether it is a one-way or two-way trust, and by controlling the issuance and usage scope of the CA agreement via naming constraint, application, and issuance policy rules.

Cross certification provides the ability to create transitive cross forest trusts, as described in the "Creating a Cross Forest Trust" section in this chapter.

Figure 1.26Figure 1.26 Hierarchical CA structure with the root (enterprise) CA at the top and subordinate (issuing) CAs below.

Editable Certificate Templates (v2)

Windows 2000 provided v1 certificate templates that were not editable. Windows 2003 provides v2 templates that are editable and include a mechanism to modify v1 templates and convert them to v2. The v2 templates facilitate cross certification and policy enforcement, which allows configuration of certificate trust relationships across forest boundaries. This could be trusts to CAs in other forests within the company or to a CA in a business partner's extranet. This also allows a CA to issue certificates to a CA in another forest and can link a policy to this cross certification. The policy can limit the issuance and usage scope of the CA cross- certification agreement.

A new Certificate Templates snap-in is included in Certificate Services for Windows Server 2003, as shown in Figure 1.27. Note that the icons for v1 and v2 templates are a different color, the Minimum Supported column identifies v2 templates as Windows 2003 and Auto Enrollment functionality is listed. This snap-in is used to modify Version 2 templates to perform the following functions:

  • Create a new template, modify an existing template, or duplicate an existing template.

  • Modify template properties such as certificate lifetime, renewal period, whether it is published in AD, Issuance Requirements such as re-enrollment, extensions, and other properties.

  • Define accounts that can enroll and autoenroll a certificate template.

  • Enable a template for Autoenrollment.

  • Set which accounts can enroll and auto-enroll for a particular certificate template. This is done by right-clicking on a template, selecting its properties and modifying the template's Access Control List (ACL) in the "security" tab. Windows.NET ACLs contain an Access Control Entry (ACE) for "Enroll" and "AutoEnroll."

Figure 1.27Figure 1.27 Certificate Template snap-in.

NOTE

You can copy and save a v1 template as a v2 template and then configure it as desired. You also can select the Duplicate Template option when you right-click on the template in the Certificate Templates snap-in, which will make a duplicate and give you a head start on creating a similar one.

Private Key Recovery

Windows Server 2003 PKI made significant changes to the key recovery process, providing centralized key recovery services. An important component to this process is the Key Recovery Agent—usually one or more Administrators who have the ability to recover encrypted files, by means of a special public key stored in an EFS Recovery Agent certificate. The key recovery data is stored in the CA database. A user's archived private key is encrypted using a symmetric key, which is encrypted using a Key Recovery Agent's public key. These keys are stored in the CA database in the RawArchivedKey and KeyRecoveryHashes columns, respectively.

To recover a user's private key, an Administrator must be added to the Recovery Agent list. The preparation for the Key Recovery process includes the following steps:

  1. Issue the EFS Recovery Agent certificate to an Administrator.

  2. Create a GPO called EFS Recovery, for example.

  3. In the GPO, go to Computer Configuration\Windows Settings\Security Settings\Public Key Policies\Encrypting File System. Define an EFS Policy. Select the Add a Recovery Agent option and add a user account that has the EFS recovery certificate (from step 2).

  4. In the Certificate Authority snap-in, in the RootCA properties, select the Recovery Agents tab. Select the Archive the Key option and select the Add button. Select the Key Recovery Agent certificate in the next dialog box.

NOTE

Multiple Administrators can be designated as Recovery Agents so that both public keys are required to recover private keys for an added measure of security.

This four-step process will recover a private key in Windows Server 2003:

  1. The Administrator who has the Recovery Agent rights (key) must know the User Principal Name (UPN) or serial number of the certificate whose private key he wants to recover.

  2. From a command prompt, run the command:

  3. certutil –getkey <serial number or UPN> <outputfile>

    This exports the recovery data from the CA database.

  4. From a command prompt, run the command:

  5. certutil –recoverkey <outputfile> <pkcs12file>

    This will convert the output file in a PKCS12 format and could store it on a floppy disk.

  6. The PKCS12 file is provided to the appropriate user, who can then import it in his certificate store.

Enhanced Security Management—Effective Permissions Tab

The Effective Permissions tab displays the cumulative permissions calculated from group membership and any inherited permissions, and displays the result applied to a given user. This is part of the Advanced Settings on the ACL properties page. For instance, right-click on an NTFS folder, go to Properties, select the Security tab, and then select the Advanced button. In the Effective Permissions tab, shown in Figure 1.28, you see a cumulative summary of all permissions applied to that user on that object.

User Autoenrollment

Windows 2000 Certificate Services provided a valuable feature in the autoenrollment and renewal of machine certificates. This allowed DCs to obtain and renew certificates. Windows Server 2003 now provides autoenrollment of users. You can do this in a Certificate Template by going to the Properties of the template, and then on the Security tab, giving Autoenroll the READ privilege. You also can enable it via Group Policy, as shown in Figure 1.29. In Computer Configuration\Windows Settings\Security Settings\Public Key Policies, check the Enroll Certificates Automatically option.

Kerberos and Forest Trusts

Windows 2000 was unable to create a transitive trust between forests, allowing only a one-way NTLM type trust that had to be created between individual domains in different forests—much like what we had to do in Windows NT to build trusts between all domains. Interestingly enough, this actually has a side benefit—because this type of trust can't be used by Kerberos, it solves a migration problem known as the "Pile On" issue, described in Chapter 3. Other than that, an NTLM trust isn't a good thing because it requires the Administrator to manage a lot of trusts between forests.

Windows Server 2003 includes schema modifications made to the trusted domain object (TDO), which allows Kerberos to authenticate across the domain. This provides transitive trusts to be built across forests. That is, we create a trust between two forests, (one way or two way) and no other trust is needed. The Administrator does have some control over the trust, in making it a one-way trust or a two-way trust, and determining the default level of authentication that is allowed between forests (described in more detail in the next section).

Figure 1.28Figure 1.28 Effective Permissions tab.

Figure 1.29Figure 1.29 Using Group Policy to enable autoenrollment of users.

Creating a Cross Forest Trust

The preparatory steps to creating a cross forest trust include establishing DNS name resolution between the two forests, synchronizing system time between all computers in both forests (easier than it sounds), and setting the forest functional level to Windows 2003 (native). Windows Server 2003 provides three different ways to establish DNS name resolution between the two forests:

  • Create a stub zone in the forest root of each forest, for the other forest.

  • Create a secondary zone in the forest root of each forest, for the other forest.

  • Set up a conditional forwarder on the DNS server that is authoritative for each root zone for the other forest.

Test name resolution by pinging the root domain name of each forest from the other forest. In addition, you must ensure that the system time of all computers in both forests is within the allowable Kerberos time skew (default is five minutes). If the time skew between the two DCs contacted to build the trust is more than the allowable time skew, the trust might be created successfully, but authentication will fail.

NOTE

There is no default time synchronization between two forests. Because Kerberos authentication, by default, requires the system time of all computers to be within five minutes of each other, ensure the time between forests is synchronized. It is recommended that you configure the PDC emulator of the root domain in each forest to synchronize with the same external time source. Time synchronization is configured by default for a forest, but not between forests. More information is available in the "Time Services" section of Chapter 6.

Setting the forest functional level is described in Chapter 3. To do this, all DCs in every domain must be Windows 2003 servers, and the domain functionality must be set to Windows Server 2003 (referred to as native, though the word "native" is not used in the UI). After the domains are all raised to Windows 2003, the forest must be raised to Windows 2003 native—similar to the way Windows 2000 domains were raised to native mode, but Windows 2000 didn't have a forest native mode.

After the forest is at Windows 2003 level, Time Services are synchronized, and DNS name resolution is established between the two forests, the trust can be created. You create the trust just as you did in Windows 2000 by going to the Active Directory Domains and Trusts snap-in, choosing Domain Properties, selecting the Trusts tab, and clicking New Trust. This engages the New Trust Wizard. Some of the options specified in the wizard are described here, along with screen shots of the dialog boxes used in the wizard.

  • Trust Type: Can create an External Trust (downlevel, nontransitive trust) between domains, or a Forest Trust (transitive, Kerberos) between forests. The dialog box is shown in Figure 1.30.

  • Figure 1.30Figure 1.30 The New Trust Wizard allows you to create an External (NTLM type) Trust or a Forest Trust (Kerberos transitive trust).

  • Trust Direction—Incoming: Users in the source domain can be authenticated in the target domain or forest (source domain is the trusted domain). The dialog box is shown in Figure 1.31.

  • Trust Direction—Outgoing: Users in the target domain or forest can be authenticated in the source domain or forest (source domain is the trusting domain).

  • Trust Direction—Two Way: Outgoing and incoming trust. Users in source and target domains or forests can be authenticated in each domain.

  • Authentication Level—Forest-wide Authentication: Users from the other forest are automatically authenticated in the local forest. Essentially this makes the authenticated users in the other forest included in the authenticated users in the local forest. This is appropriate for a company with a multiple forest deployment, and users from both forests are equally trusted for resources in each forest. They can be denied access by changing permissions just like a single forest implementation. This is defined for the ingoing and the outgoing trust separately. The dialog box is shown in Figure 1.32.

  • Figure 1.31Figure 1.31 The Trust Direction dialog box lets you specify whether the trust is incoming (trusted), outgoing (trusting), or two–way.

  • Authentication Level—Selective Authentication: This scope requires users to be specifically trusted to access any resource. This is appropriate for enterprises that have business partner extranets as separate forests and want to grant limited and specific access. This is defined for the incoming and the outgoing trust separately.

The wizard also gives you the ability to choose to verify the trust. I recommend that you do so. After the trust is created, you will be able to see users and groups of the other forest in the object picker to assign permissions to resources. For instance, in Figure 1.33, we have a share on a server in forest CorpB.Net, and in assigning permissions, the users and groups of forest CorpA.Net are available. Note that all you see in the locations option of the object picker is the remote forest with a triangle-like icon. Child domains in that forest, if any, are not exposed. However, entering user or group names (including the use of wild cards) in the object picker will find objects in any domain in the remote forest. For instance, if you entered Admin as the name in the object picker, it would return the Administrator account for all domains in the remote forest.

Figure 1.32Figure 1.32 Authentication scope options allow you to specify the default authentication between two forests.

Figure 1.33Figure 1.33 The object picker displays the trusted forest for assigning rights.

The cross forest trust will be a valuable benefit to enterprises with multiple forest deployments. Where Windows 2000's NTLM trusts have made multiple forests prohibitive to manage, Windows Server 2003's cross forest trusts could easily make the multiple forest configuration feasible.

Conclusion

This is by no means a comprehensive dissertation on security or even Windows 2003 security, but it does give a good overview of the many improvements provided in Windows Server 2003. Hopefully, after reading these issues, you will have some area you want more information on that might affect your Windows infrastructure and should be included in the migration plan. I have the good fortune of getting much of this information from HP's Jan de Clercq, a recognized security expert and author of two books regarding PKI and other security issues in the Windows environment. I highly recommend Jan's latest book, Windows Server 2003 Security Infrastructures, Digital Press, 2004.

  • + Share This
  • 🔖 Save To Your Account