Windows Server security best practices

  • Last updated on: 2016-11-11
  • Authored by: Rackspace Support

DISCLAIMER FOR MANAGED OPERATIONS CUSTOMERS

To ensure that Rackspace has access to your server when needed, we request that you do not change the following configurations as you consider security best practices:

  • When connecting to your server, Rackspace Support logs in as the user rack by using Remote Desktop Connection to the public IP address over port 3389.

  • Rebuilding existing servers or building a new server from a snapshot requires that Administrator logins are enabled and port 445 is not blocked in the Windows firewall.

If you must change these values, contact an administrator at Rackspace to make the changes in a way that does not impact our ability to provide you with Fanatical Support®.


This article provides some general security best practices to consider when you set up a Microsoft Windows server that interacts with the public Internet. Although these best practices apply to any server in general, this article specifically addresses Rackspace Public Cloud Servers running Windows.

Use local firewall rules

By default, Rackspace Public Cloud Servers do not have a firewall device. For servers that interact with the public Internet with no firewall device, the Windows firewall is the only protection between your server resources and your private data and anybody with access to an Internet connection.

Disable as many rules on the firewall as possible. Disabling rules means that fewer ports are open and listening over the public interface, which limits the server’s exposure to anyone trying to gain access to it.

For those ports that must be open, limit access to the server by whitelisting IP addresses in those specific rules. Add the IP address from your local home or office computer to the whitelist, even if your ISP provides dynamic public IP addresses that change over time. You can make changes to the firewall rules as needed from the Cloud Control Panel by logging in to the server remotely via the console and adding a new IP address.

By limiting access to the server via IP address whitelisting, you can ensure that users who need access to the server have it, but those who don’t are blocked from those open ports. The most typical ports that need to be open in the Windows firewall for web hosting on a cloud server are as follows:

Port Service
80 HTTP - IIS sites or web application
443 HTTPS Secure IIS sites or web applications with SSL

We recommend locking down the following ports via IP address whitelisting on the public interface to limit brute-force attacks or exploitation attempts against commonly named accounts or services on the server:

Port Description
3389 Remote Desktop connectivity, for logging in remotely to the server
21 FTP For the secure transfer of data between local geographic locations and the cloud server
990 FTPS (Windows) For the secure transfer of data between local geographic locations and the cloud server incorporating an SSL certificate
5000-5050 FTP Passive ports for FTP communication
1433 SQL Default port used for SQL communication
53 DNS Default port used for DNS requests

Consider what you share

Consider what data is available to others via file sharing. We do not recommend enabling Windows file sharing because the ports that are opened on the firewall (ports 445 and 139) expose the server to unwanted connection attempts.

Some customers use their servers to host back-office software such as QuickBooks, PeachTree, Microsoft Office (Outlook for Remote Desktop sessions), or other third-party software solutions. Sometimes customers want to configure mapped network drives to allow them to easily move data from their local computers to their cloud server by way of a drive letter on the local computer. However, we do not recommend this practice. Your server is only as secure as the weakest password.

Additionally, be careful about the software that you allow your users to download and install on your server. Every software package installed increases the exposure of your server to attack.

Password policy

Whether or not you have provisioned a cloud server with a hardware firewall, as previously stated, your server is only as secure as the weakest password that has access to it. Follow these tips for passwords:

  • Use strong passwords of at least 8 to 10 characters that include uppercase and lowercase letters, numbers, and special characters (such as !, #, $, and %). Assigning simple passwords can be extremely dangerous, especially for a cloud server that is available over the public Internet.

  • Set an expiration date for each user’s password. Although it is inconvenient to have to remember a new password periodically, this practice can make your data more secure.

Because our post-build automation processes depend on the default user account of Administrator, we do not recommend changing this username on your cloud servers running Windows.

Be careful who has access to the server through the Administrator account. If multiple users need admin access to the server, create multiple accounts with admin access. It’s easier to track users in the log files by looking for a specific user account than it is to try to decipher multiple log file entries under the Administrator account.

Multiple instances of Event Id 4625 in the Security Log or Event Id 1012 in the System Log can mean that someone is trying to hack into your server, because these events are related to failed login attempts.

For users logging in over Remote Desktop Connection, ensure that they are logging off the server to free any used resources instead of simply closing their RDC windows, which leaves the session open on the server.

Active Directory

We typically discourage running Active Directory on a cloud server because the only protection from intrusion is the Windows firewall and Active Directory introduces issues into a cloud server environment. Active Directory is generally better used in a dedicated server environment where servers are placed behind physical firewalls and can be connected over VPN tunneling through that firewall appliance.

Rackspace supports a VPN only if it is through a hardware firewall in a solution called RackConnect. It is easier to implement this physical firewall setup before you create servers because, at the time this article was written, the process that connects the firewall and the servers is automated during the build process. Physical firewalls are not provisioned as quickly as cloud servers and must be requested through our Hybrid teams. For more information about physical firewalls and RackConnect, see http://www.rackspace.com/cloud/hybrid/rackconnect/.

If you do install Active Directory on a cloud server, we recommend that you run two Domain Controllers in case one fails (imaging is currently unavailable for Domain Controllers). We also recommend locking down DNS to prevent DNS amplification attacks.

SQL Server instances

For servers running Microsoft SQL Server, lock down the SQL port 1433 to listen over the internal interface only, preferably listening only for connections from a list of known IP addresses of other servers needing to access SQL Server on the server. You can allow SQL port 1433 to listen over the public interface, but this rule must be limited to only the IP addresses of the computers where the developers are connecting to the databases on the server.

If you don’t limit these connections to the server, port 1433 will be exposed and outside hackers will attempt a brute-force attack on the server over this port. These types of attacks cause high network traffic, slow the server’s performance, and even bring down sites if an important account gets locked out. By limiting access to this port, these issues are mitigated before they start.

For servers running SQL Server Standard or SQL Server Web editions, we recommend configuring maintenance plans to dump the data from the live database files into flat files that can be backed up off the server and to clean up the backups so they do not fill your hard drive

Windows updates

Ensure that Windows updates are enabled, and be mindful of the state of your server – ensure that your Windows OS is patched. Patch Tuesday, which occurs on the second Tuesday of each month in North America, is the day on which Microsoft regularly releases security patches. Customers must decide how best to implement a patching strategy that keep their servers up-to-date. By default, Rackspace Cloud Servers check for updates between 2 a.m. and 4 a.m. every day.

Server backups

Set up some type of disaster recovery plan. One option that we offer is to create cloud server images nightly and write them to your Cloud Files containers with a default retention of seven days. A snapshot of the server is taken, and the image is stored in Cloud Files for use in creating new server instances or rebuilding the existing server from that image.

We also offer file-level backup through Cloud Backups. We don’t recommend backing up the entirety of the C: drive because live files that are locked cause the backup job to complete with errors. Furthermore, the Windows system files are contained in the base images provided by us or in any custom images taken of the servers, so you don’t need to back up that data on a daily basis. We do recommend backing up the C:\inetpub (IIS) directory and any other user data that needs to be backed up. Additionally, if SQL Server maintenance plans have been configured to dump the live data into flat files for backups, we recommend that you also include those directories in the backup.

Check backup jobs to ensure that they complete successfully and that the backups are valid. Create a new server instance from an image to ensure that the image is valid, and restore a file from Cloud Backups to verify that the data being backed up can be restored.

Code

The last attack surface exposed to the Internet is the code. You and your developers must ensure that the code is enforcing proper authentication and authorization. For example, a web application should not be executed with administrator-level privileges. File authorization should be carefully defined and all inputs on the application should have the best validation possible to prevent hackers from exploiting the web application and gaining control of the server.

The following sites provide information about improving ASP .Net security:

Conclusion

Depending on the use case, customers might have other more specific needs to address when leveraging our Cloud Servers product to meet their hosting needs. However, these general recommendations are a good start when considering security while creating Windows servers, cloud or otherwise.

Continue the conversation in the Rackspace Community.