Microsoft Exchange 2016 Mailbox Server Security Technical Implementation Guide

Unauthorized or malicious data changes can compromise the integrity and usefulness of the data. Automated attacks or malicious users with elevated privileges have the ability to effect change using the same mechanisms as email administrators. Auditing any changes to access mechanisms not only supports accountability and nonrepudiation for those authorized to define the environment but also enables investigation of changes made by others who may not be authorized. Note: This administrator auditing feature audits all exchange changes regardless of the user's assigned role or permissions.

Server certificates are required for many security features in Exchange; without them, the server cannot engage in many forms of secure communication. Failure to implement valid certificates makes it virtually impossible to secure Exchange's communications.

Attackers can use automated messages to determine whether a user account is active, in the office, traveling, and so on. An attacker might use this information to conduct future attacks. Verify Automatic Forwards to remote domains are disabled, except for enterprise mail that must be restricted to forward only to .mil and .gov. domains. Before enabling this setting, configure a remote domain.

A connectivity log is a record of the SMTP connection activity of the outbound message delivery queues to the destination Mailbox server, smart host, or domain. Connectivity logging is available on Hub Transport servers and Edge Transport servers. By default, connectivity logging is disabled. If events are not recorded, it may be difficult or impossible to determine the root cause of system problems or the unauthorized activities of malicious users. Note: Transport configuration settings apply to the organization/global level of the Exchange SMTP path. By checking and setting them at the Hub server, the setting will apply to both Hub and Edge roles.

Log files help establish a history of activities and can be useful in detecting attack attempts or determining tuning adjustments to improve availability. Diagnostic logging, however, characteristically produces large volumes of data and requires care in managing the logs to prevent risk of disk capacity denial-of-service conditions. Exchange diagnostic logging is divided into 29 main "services", each of which has anywhere from 2 to 26 "categories" of events to be monitored. Each category may be set to one of four levels of logging: Lowest, Low, Medium, and High, depending on how much detail is required. Higher levels of detail require more disk space to store the audit material. Diagnostic logging is intended to help administrators debug problems with their systems, not as a general-purpose auditing tool. Because the diagnostic logs collect a great deal of information, the log files may grow large very quickly. Diagnostic log levels may be raised for limited periods of time when attempting to debug relevant pieces of Exchange functionality. Once debugging has finished, diagnostic log levels should be reduced again.

Log files help establish a history of activities and can be useful in detecting attack attempts or determining tuning adjustments to improve availability. When "message tracking" is enabled, only the sender, recipients, time, and other delivery information is included by default. Information such as the subject and message body is not included. However, the absence of the message subject line can make it difficult to locate a specific message in the log unless one knows roughly what time the message was sent. To simplify searches through these logs, Exchange offers the ability to include the message "subject line" in the log files and in the Message Tracking Center display. This can make it significantly easier to locate a specific message. However, this feature creates larger log files and will contain information that may raise privacy and legal concerns. Enterprise policy should be consulted before this feature is enabled. Also, because the log files may contain sensitive information in the form of the subject line, the log files will need to be protected, commensurate with the sensitivity level, as the content may be of interest to an attacker. For these reasons, it is recommended that subject logging not be enabled during regular production operations. Instead, treat this feature as a diagnostic that can be used if needed. The tradeoff is that finding the correct message in the message tracking logs will become more difficult because the administrator will need to search using only the time the message was sent and the message’s sender. This control will have no effect unless Message Tracking is enabled. However, the setting should be disabled in case message tracking is enabled in the future.

A message tracking log provides a detailed log of all message activity as messages are transferred to and from a computer running Exchange. If events are not recorded, it may be difficult or impossible to determine the root cause of system problems or the unauthorized activities of malicious users.

Monitors are automated "process watchers" that respond to performance changes and can be useful in detecting outages and alerting administrators where attention is needed. Exchange has built-in monitors that enable the administrator to generate alerts if thresholds are reached, better enabling them to react in a timely fashion. This field offers choices of alerts when a "warning" or "critical" threshold is reached on the SMTP queue. A good rule of thumb (default) is to issue warnings when SMTP queue growth exceeds 10 minutes and critical messages when it exceeds 20 minutes, which should only happen occasionally. Frequent alerts against this counter may indicate a network or other issue (such as inbound ExchangeMER traffic) that directly impacts email delivery. Notification choices include email alert to an email-enabled account (for example, an email Administrator) or invoke a script to take other action (for example, to add an Event to the Microsoft Application Event Log, where external monitors might detect it).

It is detrimental for applications to provide, or install by default, functionality exceeding requirements or mission objectives. These unnecessary capabilities or services are often overlooked and therefore may remain unsecured. They increase the risk to the platform by providing additional attack vectors. Applications are capable of providing a wide variety of functions and services. Some of the functions and services, provided by default, may not be necessary to support essential organizational operations (e.g., key missions, functions). Examples of non-essential capabilities include but are not limited to advertising software or browser plug-ins not related to requirements or providing a wide array of functionality not required for every mission but that cannot be disabled. All system errors in Exchange will result in outbound traffic that may be identified by an eavesdropper. For this reason, the "Report Fatal Errors to Microsoft" feature must be disabled.

Log files help establish a history of activities and can be useful in detecting attack attempts or determining tuning adjustments to improve availability. Audit log content must always be considered sensitive and in need of protection. Audit data available for modification by a malicious user can be altered to conceal malicious activity. Audit data might also provide a means for the malicious user to plan unauthorized activities that exploit weaknesses. The contents of audit logs are protected against unauthorized access, modification, or deletion. Only authorized auditors and the audit functions should be granted "Read" and "Write" access to audit log data.

It is detrimental for applications to provide, or install by default, functionality exceeding requirements or mission objectives. These unnecessary capabilities or services are often overlooked and therefore may remain unsecured. They increase the risk to the platform by providing additional attack vectors. Applications are capable of providing a wide variety of functions and services. Some of the functions and services, provided by default, may not be necessary to support essential organizational operations (e.g., key missions, functions). Examples of non-essential capabilities include, but are not limited to, advertising software or browser plug-ins not related to requirements or providing a wide array of functionality not required for every mission but that cannot be disabled. Customer Experience reports in Exchange will result in outbound traffic that may be identified by an eavesdropper. For this reason, the Customer Experience reports must not be sent to Microsoft.

Log files help establish a history of activities and can be useful in detecting attack attempts or determining tuning adjustments to improve availability. Audit log content must always be considered sensitive and in need of protection. Audit data available for modification by a malicious user can be altered to conceal malicious activity. Audit data might also provide a means for the malicious user to plan unauthorized activities that exploit weaknesses. The contents of audit logs are protected against unauthorized access, modification, or deletion. Only authorized auditors and the audit functions should be granted "Read" and "Write" access to audit log data.

Log files help establish a history of activities and can be useful in detecting attack attempts or determining tuning adjustments to improve availability. Audit log content must always be considered sensitive and in need of protection. Audit data available for modification by a malicious user can be altered to conceal malicious activity. Audit data might also provide a means for the malicious user to plan unauthorized activities that exploit weaknesses. The contents of audit logs are protected against unauthorized access, modification, or deletion. Only authorized auditors and the audit functions should be granted "Read" and "Write" access to audit log data.

Log files help establish a history of activities and can be useful in detecting attack attempts or determining tuning adjustments to improve availability. Audit log content must always be considered sensitive and in need of protection. Successful exploit of an application server vulnerability may well be logged by monitoring or audit processes when it occurs. Writing log and audit data to a separate partition where separate security contexts protect them may offer the ability to protect this information from being modified or removed by the exploit mechanism.

Scripts often provide a way for attackers to infiltrate a system, especially scripts downloaded from untrusted locations. By setting machine policy to prevent unauthorized script executions, unanticipated system impacts can be avoided. Failure to allow only signed remote scripts reduces the attack vector vulnerabilities from unsigned remote scripts.

IMAP4 is not approved for use within the DoD. It uses a clear-text-based user name and password and does not support the DoD standard for PKI for email access. User name and password could easily be captured from the network, allowing a malicious user to access other system features. Uninstalling or disabling the service will prevent the use of the IMAP4 protocol.

POP3 is not approved for use within the DoD. It uses a clear-text-based user name and password and does not support the DoD standard for PKI for email access. User name and password could easily be captured from the network, allowing a malicious user to access other system features. Uninstalling or disabling the service will prevent the use of POP3.

In the same way that added security layers can provide a cumulative positive effect on security posture, multiple applications can provide a cumulative negative effect. A vulnerability and subsequent exploit to one application can lead to an exploit of other applications sharing the same security context. For example, an exploit to a web server process that leads to unauthorized administrative access to the host system can most likely lead to a compromise of all applications hosted by the same system. Email services should be installed to a discrete set of directories on a partition that does not host other applications. Email services should never be installed on a Domain Controller/Directory Services server.

When identifying a "Smart Host" for the email environment, a logical Send connector is the preferred method. A Smart Host acts as an Internet-facing concentrator for other email servers. Appropriate hardening can be applied to the Smart Host, rather than at multiple locations throughout the enterprise. Failure to identify a Smart Host could default to each email server performing its own lookups (potentially through protective firewalls). Exchange servers should not be Internet facing and should therefore not perform any Smart Host functions. When the Exchange servers are Internet facing, they must be configured to identify the Internet-facing server that is performing the Smart Host function.

The Simple Mail Transfer Protocol (SMTP) Receive connector is used by Exchange to send and receive messages from server to server using SMTP protocol. This setting controls the encryption strength used for client connections to the SMTP Receive connector. With this feature enabled, only clients capable of supporting secure communications will be able to send mail using this SMTP server. Where secure channels are required, encryption can also be selected. The use of secure communication prevents eavesdroppers from reading or modifying communications between mail clients and servers. While sensitive message bodies should be encrypted by the sender at the client, requiring a secure connection from the client to the server adds protection by encrypting the sender and recipient information that cannot be encrypted by the sender. Individually, channel security and encryption have been compromised by attackers. Used together, email becomes a more difficult target, and security is heightened. Failure to enable this feature gives eavesdroppers an opportunity to read or modify messages between the client and server. Multiple values can be separated by commas, but some values have dependencies and exclusions. AuthMechanism may include other mechanisms as long as the "Tls" is identified. • Only use the value "None" by itself. • The value "BasicAuthRequireTLS" requires the values "BasicAuth" and "Tls". • The only other value that can be used with ExternalAuthoritative is "Tls". • The value "Tls" is required when the value of the RequireTLS parameter is "$true". • The value "ExternalAuthoritative" requires the value of the PermissionGroups parameter be set to "ExchangeServers".

Backup and recovery procedures are an important part of overall system availability and integrity. Complete backups reduce the chance of accidental deletion of important information and make it possible to have complete recoveries. It is not uncommon for users to receive and delete messages in the scope of a single backup cycle. This setting ensures at least one backup has been run on the mailbox store before the message physically disappears. By enabling this setting, all messages written to recipients who have accounts on this store will reside in backups even if they have been deleted by the user before the backup has run.

Auto-forwarded email accounts do not meet the requirement for digital signature and encryption of Controlled Unclassified Information (CUI) and Personally Identifiable Information (PII) in accordance with DoDI 8520.2 (reference ee) and DoD Director for Administration and Management memorandum, "Safeguarding Against and Responding to the Breach of Personally Identifiable Information". Use of forwarding set by an administrator interferes with nonrepudiation requirements that each end user be responsible for creation and destination of email data.

Auto-forwarded email accounts do not meet the requirement for digital signature and encryption of Controlled Unclassified Information (CUI) and Personally Identifiable Information (PII) in accordance with DoDI 8520.2 (reference ee) and DoD Director for Administration and Management memorandum, "Safeguarding Against and Responding to the Breach of Personally Identifiable Information". Use of forwarding set by an administrator interferes with nonrepudiation requirements that each end user be responsible for creation and destination of email data.

This control is used to limit the servers that may use this server as a relay. If a Simple Mail Transport Protocol (SMTP) sender does not have a direct connection to the Internet (for example, an application that produces reports to be emailed), it will need to use an SMTP Receive connector that does have a path to the Internet (for example, a local email server) as a relay. SMTP relay functions must be protected so third parties are not able to hijack a relay service for their own purposes. Most commonly, hijacking of relays is done by spammers to disguise the source of their messages and may also be used to cover the source of more destructive attacks. Relays can be restricted in one of three ways: by blocking relays (restrict to a blank list of servers), by restricting use to lists of valid servers, or by restricting use to servers that can authenticate. Because authenticated connections are the most secure for SMTP Receive connectors, it is recommended that relays allow only servers that can authenticate.

Spam originators, in an effort to refine mailing lists, sometimes monitor transmissions for automated bounce-back messages. Automated messages include such items as "Out of Office" responses, nondelivery messages, and automated message forwarding. Automated bounce-back messages can be used by a third party to determine if users exist on the server. This can result in the disclosure of active user accounts to third parties, paving the way for possible future attacks.

Originators of spam messages are constantly changing their techniques in order to defeat spam countermeasures; therefore, spam software must be constantly updated to address the changing threat. A manual update procedure is labor intensive and does not scale well in an enterprise environment. This risk may be mitigated by using an automatic update capability. Spam protection mechanisms include, for example, signature definitions, rule sets, and algorithms. Exchange 2016 provides both anti-spam and anti-malware protection out of the box. The Exchange 2016 anti-spam and anti-malware product capabilities are limited but still provide some protection.

Originators of spam messages are constantly changing their techniques in order to defeat spam countermeasures; therefore, spam software must be constantly updated to address the changing threat. A manual update procedure is labor intensive and does not scale well in an enterprise environment. This risk may be mitigated by using an automatic update capability. Spam protection mechanisms include, for example, signature definitions, rule sets, and algorithms. Exchange 2016 provides both anti-spam and anti-malware protection out of the box. The Exchange 2016 anti-spam and anti-malware product capabilities are limited but still provide some protection.

Originators of spam messages are constantly changing their techniques in order to defeat spam countermeasures; therefore, spam software must be constantly updated to address the changing threat. A manual update procedure is labor intensive and does not scale well in an enterprise environment. This risk may be mitigated by using an automatic update capability. Spam protection mechanisms include, for example, signature definitions, rule sets, and algorithms. Exchange 2016 provides both anti-spam and anti-malware protection out of the box. The Exchange 2016 anti-spam and anti-malware product capabilities are limited but still provide some protection.

Attackers can use automated messages to determine whether a user account is active, in the office, traveling, and so on. An attacker might use this information to conduct future attacks. Remote users will not receive automated "Out of Office" delivery reports. This setting can be used to determine if all the servers in the organization can send "Out of Office" messages.

Default product installations may provide more generous access permissions than are necessary to run the application. By examining and tailoring access permissions to more closely provide the least amount of privilege possible, attack vectors that align with user permissions are less likely to access more highly secured areas.

Exchange software, as with other application software installed on a host system, must be included in a system baseline record and periodically reviewed; otherwise, unauthorized changes to the software may not be discovered. This effort is a vital step to securing the host and the applications, as it is the only method that may provide the ability to detect and recover from otherwise undetected changes, such as those that result from worm or bot intrusions. The Exchange software and configuration baseline is created and maintained for comparison during scanning efforts. Operational procedures must include baseline updates as part of configuration management tasks that change the software and configuration.

Monitoring software files for changes against a baseline on a regular basis may help detect the possible introduction of malicious code on a system.

Unneeded but running services offer attackers an enhanced attack profile, and attackers are constantly watching to discover open ports with running services. By analyzing and disabling unneeded services, the associated open ports become unresponsive to outside queries, and servers become more secure as a result. Exchange Server has role-based server deployment to enable protocol path control and logical separation of network traffic types. For example, a server implemented in the Client Access role (i.e., Outlook Web App [OWA]) is configured and tuned as a web server using web protocols. A client access server exposes only web protocols (HTTP/HTTPS), enabling system administrators to optimize the protocol path and disable all services unnecessary for Exchange web services. Similarly, servers created to host mailboxes are dedicated to that task and must operate only the services needed for mailbox hosting. (Exchange servers must also operate some web services, but only to the degree that Exchange requires the IIS engine in order to function). Because Post Office Protocol 3 (POP3) and Internet Message Access Protocol 4 (IMAP4) clients are not included in the standard desktop offering, they must be disabled. While IMAP4 is restricted, IMAP Secure is not restricted and does not apply to this requirement.

Identification and authentication provide the foundation for access control. Access to email services applications require NTLM authentication. Outlook Anywhere, if authorized for use by the site, must use NTLM authentication when accessing email. Note: There is a technical restriction in Exchange Outlook Anywhere that requires a direct SSL connection from Outlook to the Certificate Authority (CA) server. There is also a constraint where Microsoft supports that the CA server must participate in the Active Director (AD) domain inside the enclave. For this reason, Outlook Anywhere must be deployed only for enclave-sourced Outlook users.

In the same way that added security layers can provide a cumulative positive effect on security posture, multiple applications can provide a cumulative negative effect. A vulnerability and subsequent exploit to one application can lead to an exploit of other applications sharing the same security context. For example, an exploit to a web server process that leads to unauthorized administrative access to the host system can most likely lead to a compromise of all applications hosted by the same system. Email services should be installed on a partition that does not host other applications. Email services should never be installed on a Domain Controller/Directory Services server.

Attackers can use automated messages to determine whether a user account is active, in the office, traveling, and so on. An attacker might use this information to conduct future attacks. Ensure that delivery reports to remote domains are disabled. Before enabling this setting, first configure a remote domain using the Exchange Management Console (EMC) or the New-RemoteDomain cmdlet.

Attackers can use automated messages to determine whether a user account is active, in the office, traveling, and so on. An attacker might use this information to conduct future attacks. Ensure that nondelivery reports to remote domains are disabled. Before enabling this setting, first configure a remote domain using the Exchange Management Console (EMC) or the New-RemoteDomain cmdlet.

Automated connection responses occur as a result of FTP or Telnet connections when connecting to those services. They report a successful connection by greeting the connecting client and stating the name, release level, and (often) additional information regarding the responding product. While useful to the connecting client, connection responses can also be used by a third party to determine operating system or product release levels on the target server. The result can include disclosure of configuration information to third parties, paving the way for possible future attacks. For example, when querying the SMTP service on port 25, the default response looks similar to this one: 220 exchange.mydomain.org Microsoft ESMTP MAIL Service, Version: 6.0.3790.211 ready at Wed, 2 Feb 2005 23:40:00 -0500 Changing the response to hide local configuration details reduces the attack profile of the target.

The Simple Mail Transfer Protocol (SMTP) connector is used by Exchange to send and receive messages from server to server. Several controls work together to provide security between internal servers. This setting controls the encryption method used for communications between servers. With this feature enabled, only servers capable of supporting Transport Layer Security (TLS) will be able to send and receive mail within the domain. The use of secure communication prevents eavesdroppers from reading or modifying communications between mail clients and servers. While sensitive message bodies should be encrypted by the sender at the client, requiring a secure connection from server to server adds protection by encrypting the sender and recipient information that cannot be encrypted by the sender. Individually, channel security and encryption can be compromised by attackers. Used together, email becomes a more difficult target, and security is heightened. Failure to enable this feature gives eavesdroppers an opportunity to read or modify messages between servers.

Exchange Server mailbox databases and any data contained in those mailboxes should be protected. This can be accomplished by configuring Mailbox servers and databases for high availability and site resilience. A database availability group (DAG) is a component of the Mailbox server high availability and site resilience framework built into Microsoft Exchange Server 2016. A DAG is a group of Mailbox servers that hosts a set of databases and provides automatic database-level recovery from failures that affect individual servers or databases. A DAG is a boundary for mailbox database replication and database and server switchovers and failovers. Any server in a DAG can host a copy of a mailbox database from any other server in the DAG. When a server is added to a DAG, it works with the other servers in the DAG to provide automatic recovery from failures that affect mailbox databases, such as a disk, server, or network failure.

Failure to install the most current Exchange service pack leaves a system vulnerable to exploitation. Current service packs correct known security and system vulnerabilities.

Configuring the application to implement organization-wide security implementation guides and security checklists ensures compliance with Federal standards and establishes a common security baseline across DoD that reflects the most restrictive security posture consistent with operational requirements. Configuration settings are the set of parameters that can be changed that affect the security posture and/or functionality of the system. Security-related parameters are those parameters impacting the security state of the application, including the parameters required to satisfy other security control requirements.

Malicious code protection mechanisms include but are not limited to anti-virus and malware detection software. To minimize potential negative impact to the organization that can be caused by malicious code, it is imperative that malicious code is identified and eradicated. Malicious code includes viruses, worms, trojan horses, and spyware. It is not enough to have the software installed; this software must periodically scan the system to search for malware on an organization-defined frequency. Exchange's built-in Malware Agent is not designed to address all malicious code protection workloads. This workload is best handled by third-party anti-virus and intrusion prevention software. Sites must use an approved DoD scanner. Exchange Malware software has a limited scanning capability and does not scan files that are downloaded, opened, or executed.

This setting controls whether client machines are forced to use secure channels to communicate with the server. If this feature is enabled, clients will only be able to communicate with the server over secure communication channels. Failure to require secure connections to the client access server increases the potential for unintended eavesdropping or data loss.

This setting controls whether client machines should be forced to use secure channels to communicate with this virtual directory. If this feature is enabled, clients will only be able to communicate with the directory if they are capable of supporting secure communication with the server. The use of secure communication prevents eavesdroppers from reading or modifying communications between servers and clients. The network and DMZ STIG identify criteria for OWA and Public Folder configuration in the network, including Common Access Card (CAC)-enabled pre-authentication through an application firewall proxy. Failure to require secure connections on a website increases the potential for unintended eavesdropping or data loss.

Identification and Authentication provide the foundation for access control. Access to email services applications in the DoD requires authentication using DoD Public Key Infrastructure (PKI) certificates. Authentication for Outlook Web App (OWA) is used to enable web access to user email mailboxes and should assume that certificate-based authentication has been configured. This setting controls whether forms-based logon should be used by the OWA website. Because the DoD requires Common Access Card (CAC)-based authentication to applications, OWA access must be brokered through an application proxy or other pre-authenticator, which performs CAC authentication prior to arrival at the CA server. The authenticated request is then forwarded directly to OWA, where authentication is repeated without requiring the user to repeat authentication steps. For this scenario to work, the Application Proxy server must have forms-based authentication enabled, and Exchange must have forms-based Authentication disabled. If forms-based Authentication is enabled on the Exchange CA server, it is evidence that the application proxy server is either not correctly configured, or it may be missing.

To mitigate the risk of unauthorized access to sensitive information by entities that have been issued certificates by DoD-approved PKIs, all DoD systems (e.g., networks, web servers, and web portals) must be properly configured to incorporate access control methods that do not rely solely on the possession of a certificate for access. Successful authentication must not automatically give an entity access to an asset or security boundary. Authorization procedures and controls must be implemented to ensure each authenticated entity also has a validated and current authorization. Authorization is the process of determining whether an entity, once authenticated, is permitted to access a specific asset. Information systems use access control policies and enforcement mechanisms to implement this requirement. Access control policies include identity-based policies, role-based policies, and attribute-based policies. Access enforcement mechanisms include access control lists, access control matrices, and cryptography. These policies and mechanisms must be employed by the application to control access between users (or processes acting on behalf of users) and objects (e.g., devices, files, records, processes, programs, and domains) in the information system. This requirement is applicable to access control enforcement applications (e.g., authentication servers) and other applications that perform information and system access control functions.