VMware vSphere 8.0 ESXi Security Technical Implementation Guide

By limiting the number of failed logon attempts, the risk of unauthorized access via user password guessing, otherwise known as brute forcing, is reduced. Once the configured number of attempts is reached, the account is locked by the ESXi host.

Display of a standardized and approved use notification before granting access to the host ensures privacy and security notification verbiage used is consistent with applicable federal laws, Executive Orders, directives, policies, regulations, standards, and guidance. System use notifications are required only for access via logon interfaces with human users and are not required when such human interfaces do not exist. The banner must be formatted in accordance with applicable DOD policy. Use the following verbiage for a host that can accommodate banners of 1300 characters: "You are accessing a U.S. Government (USG) VMM (IS) that is provided for USG-authorized use only. By using this IS (which includes any device attached to this IS), you consent to the following conditions: -The USG routinely intercepts and monitors communications on this IS for purposes including, but not limited to, penetration testing, COMSEC monitoring, network operations and defense, personnel misconduct (PM), law enforcement (LE), and counterintelligence (CI) investigations. -At any time, the USG may inspect and seize data stored on this IS. -Communications using, or data stored on, this IS are not private, are subject to routine monitoring, interception, and search, and may be disclosed or used for any USG-authorized purpose. -This IS includes security measures (e.g., authentication and access controls) to protect USG interests--not for your personal benefit or privacy. -Notwithstanding the above, using this IS does not constitute consent to PM, LE or CI investigative searching or monitoring of the content of privileged communications, or work product, related to personal representation or services by attorneys, psychotherapists, or clergy, and their assistants. Such communications and work product are private and confidential. See User Agreement for details." Use the following verbiage for VMMs that have severe limitations on the number of characters that can be displayed in the banner: "I've read (literal ampersand) consent to terms in IS user agreem't." Satisfies: SRG-OS-000023-VMM-000060, SRG-OS-000024-VMM-000070

Enabling Lockdown Mode disables direct access to an ESXi host, requiring the host to be managed remotely from vCenter Server. This is done to ensure the roles and access controls implemented in vCenter are always enforced and users cannot bypass them by logging on to a host directly. By forcing all interaction to occur through vCenter Server, the risk of someone inadvertently attaining elevated privileges or performing tasks that are not properly audited is greatly reduced.

The ESXi host client is the UI served up by the host itself, outside of vCenter. It is accessed at https:///ui. ESXi is not usually administered via this interface for long periods, and all users will be highly privileged. Implementing a mandatory session idle limit will ensure that orphaned, forgotten, or ignored sessions will be closed promptly.

Without establishing what types of events occurred, it would be difficult to establish, correlate, and investigate the events leading up to an outage or attack. Satisfies: SRG-OS-000037-VMM-000150, SRG-OS-000063-VMM-000310

To enforce the use of complex passwords, minimum numbers of characters of different classes are mandated. The use of complex passwords reduces the ability of attackers to successfully obtain valid passwords using guessing or exhaustive search techniques. Complexity requirements increase the password search space by requiring users to construct passwords from a larger character set than they may otherwise use. Satisfies: SRG-OS-000069-VMM-000360, SRG-OS-000070-VMM-000370, SRG-OS-000071-VMM-000380, SRG-OS-000072-VMM-000390, SRG-OS-000072-VMM-000390, SRG-OS-000078-VMM-000450, SRG-OS-000266-VMM-000940

If a user or root used the same password continuously or was allowed to change it back shortly after being forced to change it to something else, it would provide a potential intruder with the opportunity to keep guessing at one user's password until it was guessed correctly.

The MOB provides a way to explore the object model used by the VMkernel to manage the host and enables configurations to be changed. This interface is meant to be used primarily for debugging the vSphere Software Development Kit (SDK), but because there are no access controls it could also be used as a method to obtain information about a host being targeted for unauthorized access.

SSH trust relationships mean a compromise on one host can allow an attacker to move trivially to other hosts. SSH can emulate the behavior of the obsolete "rsh" command in allowing users to enable insecure access to their accounts via ".rhosts" files.

If a user forgets to log out of their local or remote ESXi Shell session, the idle connection will remain open indefinitely and increase the likelihood of inappropriate host access via session hijacking. The "ESXiShellInteractiveTimeOut" allows the automatic termination of idle shell sessions. Satisfies: SRG-OS-000163-VMM-000700, SRG-OS-000279-VMM-001010

Secure Boot is part of the UEFI firmware standard. With UEFI Secure Boot enabled, a host refuses to load any UEFI driver or app unless the operating system bootloader has a valid digital signature. Secure Boot for ESXi requires support from the firmware and it requires that all ESXi kernel modules, drivers, and VIBs be signed by VMware or a partner subordinate. Secure Boot is enabled in the BIOS of the ESXi physical server and supported by the hypervisor boot loader. This control flips ESXi from merely supporting Secure Boot to requiring it. Without this setting enabled, and configuration encryption, an ESXi host could be subject to offline attacks. An attacker could simply transfer the ESXi install drive to a non-Secure Boot host and boot it up without ESXi complaining. Satisfies: SRG-OS-000257-VMM-000910, SRG-OS-000258-VMM-000920, SRG-OS-000445-VMM-001780, SRG-OS-000446-VMM-001790

Secure Boot is part of the Unified Extensible Firmware Interface (UEFI) firmware standard. With UEFI Secure Boot enabled, a host refuses to load any UEFI driver or app unless the operating system bootloader has a valid digital signature. Secure Boot for ESXi requires support from the firmware and requires that all ESXi kernel modules, drivers, and vSphere Installation Bundles (VIBs) be signed by VMware or a partner subordinate. Secure Boot is enabled in the BIOS of the ESXi physical server and supported by the hypervisor boot loader. There is no ESXi control to "turn on" Secure Boot. Requiring Secure Boot (failing to boot without it present) is accomplished in another control.

By enforcing a reasonable unlock timeout after multiple failed logon attempts, the risk of unauthorized access via user password guessing, otherwise known as brute forcing, is reduced. Users must wait for the timeout period to elapse before subsequent logon attempts are allowed.

In order to ensure ESXi has sufficient storage capacity in which to write the audit logs, audit record storage capacity should be configured. If a central audit record storage facility is available, the local storage capacity should be sufficient to hold audit records that would accumulate during anticipated interruptions in delivery of records to the facility.

Remote logging to a central log host provides a secure, centralized store for ESXi logs. By gathering host log files onto a central host, it can more easily monitor all hosts with a single tool. It can also do aggregate analysis and searching to look for such things as coordinated attacks on multiple hosts. Logging to a secure, centralized log server also helps prevent log tampering and provides a long-term audit record. Satisfies: SRG-OS-000342-VMM-001230, SRG-OS-000274-VMM-000960, SRG-OS-000275-VMM-000970, SRG-OS-000277-VMM-000990, SRG-OS-000479-VMM-001990

To ensure the accuracy of the system clock, it must be synchronized with an authoritative time source within DOD. Many system functions, including time-based logon and activity restrictions, automated reports, system logs, and audit records, depend on an accurate system clock. If there is no confidence in the correctness of the system clock, time-based functions may not operate as intended and records may be of diminished value. Satisfies: SRG-OS-000355-VMM-001330, SRG-OS-000356-VMM-001340

When enabled, vSphere performs bidirectional authentication of both the iSCSI target and host. When not authenticating both the iSCSI target and host, there is potential for a man-in-the-middle attack, in which an attacker might impersonate either side of the connection to steal data. Bidirectional authentication mitigates this risk.

While encrypted vMotion is available, vMotion traffic should still be sequestered from other traffic to further protect it from attack. This network must only be accessible to other ESXi hosts, preventing outside access to the network. The vMotion VMkernel port group must be in a dedicated VLAN that can be on a standard or distributed virtual switch as long as the vMotion VLAN is not shared by any other function and is only routed to ESXi hosts.

Use of weak or untested encryption algorithms undermines the purposes of utilizing encryption to protect data. ESXi must implement cryptographic modules adhering to the higher standards approved by the federal government since this provides assurance they have been tested and validated.

Lockdown mode disables direct host access, requiring that administrators manage hosts from vCenter Server. However, if a host becomes isolated from vCenter, the administrator is locked out and can no longer manage the host. The "DCUI.Access" advanced setting allows specified users to exit lockdown mode in such a scenario. If the Direct Console User Interface (DCUI) is running in strict lockdown mode, this setting is ineffective.

Display of a standardized and approved use notification before granting access to the host ensures privacy and security notification verbiage used is consistent with applicable federal laws, Executive Orders, directives, policies, regulations, standards, and guidance. System use notifications are required only for access via logon interfaces with human users and are not required when such human interfaces do not exist. The banner must be formatted in accordance with applicable DOD policy. Use the following verbiage for a host that can accommodate banners of 1300 characters: "You are accessing a U.S. Government (USG) VMM (IS) that is provided for USG-authorized use only. By using this IS (which includes any device attached to this IS), you consent to the following conditions: -The USG routinely intercepts and monitors communications on this IS for purposes including, but not limited to, penetration testing, COMSEC monitoring, network operations and defense, personnel misconduct (PM), law enforcement (LE), and counterintelligence (CI) investigations. -At any time, the USG may inspect and seize data stored on this IS. -Communications using, or data stored on, this IS are not private, are subject to routine monitoring, interception, and search, and may be disclosed or used for any USG-authorized purpose. -This IS includes security measures (e.g., authentication and access controls) to protect USG interests--not for your personal benefit or privacy. -Notwithstanding the above, using this IS does not constitute consent to PM, LE or CI investigative searching or monitoring of the content of privileged communications, or work product, related to personal representation or services by attorneys, psychotherapists, or clergy, and their assistants. Such communications and work product are private and confidential. See User Agreement for details." Use the following verbiage for VMMs that have severe limitations on the number of characters that can be displayed in the banner: "I've read (literal ampersand) consent to terms in IS user agreem't."

Display of a standardized and approved use notification before granting access to the host ensures privacy and security notification verbiage used is consistent with applicable federal laws, Executive Orders, directives, policies, regulations, standards, and guidance. System use notifications are required only for access via logon interfaces with human users and are not required when such human interfaces do not exist. The banner must be formatted in accordance with applicable DOD policy. Use the following verbiage for a host that can accommodate banners of 1300 characters: "You are accessing a U.S. Government (USG) VMM (IS) that is provided for USG-authorized use only. By using this IS (which includes any device attached to this IS), you consent to the following conditions: -The USG routinely intercepts and monitors communications on this IS for purposes including, but not limited to, penetration testing, COMSEC monitoring, network operations and defense, personnel misconduct (PM), law enforcement (LE), and counterintelligence (CI) investigations. -At any time, the USG may inspect and seize data stored on this IS. -Communications using, or data stored on, this IS are not private, are subject to routine monitoring, interception, and search, and may be disclosed or used for any USG-authorized purpose. -This IS includes security measures (e.g., authentication and access controls) to protect USG interests--not for your personal benefit or privacy. -Notwithstanding the above, using this IS does not constitute consent to PM, LE or CI investigative searching or monitoring of the content of privileged communications, or work product, related to personal representation or services by attorneys, psychotherapists, or clergy, and their assistants. Such communications and work product are private and confidential. See User Agreement for details." Use the following verbiage for VMMs that have severe limitations on the number of characters that can be displayed in the banner: "I've read (literal ampersand) consent to terms in IS user agreem't."

The ESXi Shell is an interactive command line interface (CLI) available at the ESXi server console. The ESXi shell provides temporary access to commands essential for server maintenance. Intended primarily for use in break-fix scenarios, the ESXi shell is well suited for checking and modifying configuration details, which are not always generally accessible, using the vSphere Client. The ESXi shell is accessible remotely using SSH by users with the Administrator role. Under normal operating conditions, SSH access to the host must be disabled as is the default. As with the ESXi shell, SSH is also intended only for temporary use during break-fix scenarios. SSH must therefore be disabled under normal operating conditions and must only be enabled for diagnostics or troubleshooting. Remote access to the host must therefore be limited to the vSphere Client or Host Client at all other times. Satisfies: SRG-OS-000095-VMM-000480, SRG-OS-000297-VMM-001040, SRG-OS-000298-VMM-001050

The ESXi Shell is an interactive command line environment available locally from the Direct Console User Interface (DCUI) or remotely via SSH. Activities performed from the ESXi Shell bypass vCenter role-based access control (RBAC) and audit controls. The ESXi shell must only be turned on when needed to troubleshoot/resolve problems that cannot be fixed through the vSphere client.

When the ESXi Shell or Secure Shell (SSH) services are enabled on a host, they will run indefinitely. To avoid having these services left running, set the "ESXiShellTimeOut". The "ESXiShellTimeOut" defines a window of time after which the ESXi Shell and SSH services will be stopped automatically.

When the Direct Console User Interface (DCUI) is enabled and logged in, it should be automatically logged out if left logged on to avoid access by unauthorized persons. The "DcuiTimeOut" setting defines a window of time after which the DCUI will be logged out.

The vSphere management network provides access to the vSphere management interface on each component. Services running on the management interface provide an opportunity for an attacker to gain privileged access to the systems. Any remote attack most likely would begin with gaining entry to this network. The Management VMkernel port group can be on a standard or distributed virtual switch but must be on a dedicated VLAN. The Management VLAN must not be shared by any other function and must not be accessible to anything other than management-related functions such as vCenter.

Virtual machines (VMs) might share virtual switches and VLANs with the IP-based storage configurations. IP-based storage includes vSAN, iSCSI, and NFS. This configuration might expose IP-based storage traffic to unauthorized VM users. IP-based storage frequently is not encrypted. It can be viewed by anyone with access to this network. To restrict unauthorized users from viewing the IP-based storage traffic, the IP-based storage network must be logically separated from any other traffic. Configuring the IP-based storage adaptors on separate VLANs or network segments from other VMkernels and VMs will limit unauthorized users from viewing the traffic.

While a host is in lockdown mode (strict or normal), only users on the "Exception Users" list are allowed access. These users do not lose their permissions when the host enters lockdown mode. The organization may want to add service accounts such as a backup agent to the Exception Users list. Verify the list of users exempted from losing permissions is legitimate and as needed per the environment. Adding unnecessary users to the exception list defeats the purpose of lockdown mode.

SSH trust relationships mean a compromise on one host can allow an attacker to move trivially to other hosts. SSH's cryptographic host-based authentication is more secure than ".rhosts" authentication since hosts are cryptographically authenticated. However, it is not recommended that hosts unilaterally trust one another, even within an organization.

SSH environment options potentially allow users to bypass access restriction in some configurations. Users must not be able to present environment options to the SSH daemon.

OpenSSH has the ability to create network tunnels (layer 2 and layer 3) over an SSH connection. This function can provide similar convenience to a virtual private network (VPN) with the similar risk of providing a path to circumvent firewalls and network Access Control Lists (ACLs).

If SNMP is not being used, it must remain disabled. If it is being used, the proper trap destination must be configured. If SNMP is not properly configured, monitoring information can be sent to a malicious host that can use this information to plan an attack.

In addition to service-specific firewall rules, ESXi has a default firewall rule policy to allow or deny incoming and outgoing traffic. Reduce the risk of attack by ensuring this is set to deny incoming and outgoing traffic.

BPDU Guard and Portfast are commonly enabled on the physical switch to which the ESXi host is directly connected to reduce the Spanning Tree Protocol (STP) convergence delay. If a BPDU packet is sent from a virtual machine (VM) on the ESXi host to the physical switch configured as stated above, a cascading lockout of all the uplink interfaces from the ESXi host can occur. To prevent this type of lockout, BPDU Filter can be enabled on the ESXi host to drop any BPDU packets being sent to the physical switch. The caveat is that certain Secure Socket Layer (SSL) virtual private networks that use Windows bridging capability can legitimately generate BPDU packets. The administrator should verify no legitimate BPDU packets are generated by VMs on the ESXi host prior to enabling BPDU Filter. If BPDU Filter is enabled in this situation, enabling Reject Forged Transmits on the virtual switch port group adds protection against Spanning Tree loops.

If the virtual machine (VM) operating system changes the Media Access Control (MAC) address, the operating system can send frames with an impersonated source MAC address at any time. This allows an operating system to stage malicious attacks on the devices in a network by impersonating a network adaptor authorized by the receiving network. This means the virtual switch does not compare the source and effective MAC addresses. To protect against MAC address impersonation, all virtual switches must have forged transmissions set to reject. Reject Forged Transmit can be set at the vSwitch and/or the Portgroup level. Switch-level settings can be overridden at the Portgroup level.

When promiscuous mode is enabled for a virtual switch, all virtual machines (VMs) connected to the Portgroup have the potential to read all packets across that network (only the virtual machines connected to that Portgroup). Promiscuous mode is disabled by default on the ESXi Server, and this is the recommended setting. Promiscuous mode can be set at the vSwitch and/or the Portgroup level. Switch-level settings can be overridden at the Portgroup level.

If the organization is not using products that use the dvFilter network API, the host should not be configured to send network information to a virtual machine (VM). If the API is enabled, an attacker might attempt to connect a virtual machine to it, potentially providing access to the network of other VMs on the host. If using a product that makes use of this API, verify the host has been configured correctly. If not using such a product, ensure the setting is blank.

When a port group is set to VLAN 4095, the vSwitch passes all network frames to the attached virtual machines (VMs) without modifying the VLAN tags. In vSphere, this is referred to as VGT. The VM must process the VLAN information itself via an 802.1Q driver in the operating system. VLAN 4095 must only be implemented if the attached VMs have been specifically authorized and are capable of managing VLAN tags themselves. If VLAN 4095 is enabled inappropriately, it may cause denial of service or allow a VM to interact with traffic on an unauthorized VLAN.

Warnings that local or remote shell sessions are enabled alert administrators to activity they may not be aware of and need to investigate.

The L1 Terminal Fault (L1TF) CPU vulnerabilities published in 2018 have patches and mitigations available in vSphere. However, there are performance impacts to these mitigations that require careful thought and planning from the system administrator before implementation. Until a mitigation is implemented, the UI warning about the lack of a mitigation must not be dismissed so the system administrator does not assume the vulnerability has been addressed.

When sending syslog data to a remote host, ESXi can be configured to use any combination of TCP, UDP, and SSL transports. When using SSL, the server certificate must be validated to ensure that the host is connecting to a valid syslog server.

By default, pages allocated for virtual machines (VMs), userspace applications, and kernel threads are zeroed out at allocation time. ESXi will always ensure that no nonzero pages are exposed to VMs or userspace applications. While this prevents exposing cryptographic keys from VMs or userworlds to other clients, these keys can stay present in host memory for a long time if the memory is not reused. The NIAP Virtualization Protection Profile and Server Virtualization Extended Package require that memory that may contain cryptographic keys be zeroed upon process exit. To this end, a new configuration option, MemEagerZero, can be configured to enforce zeroing out userworld and guest memory pages when a userworld process or guest exits. For kernel threads, memory spaces holding keys are zeroed out as soon as the secret is no longer needed.

The vSphere API (VIM) allows for remote, programmatic administration of the ESXi host. Authenticated API sessions are no different from a risk perspective than authenticated UI sessions and they need similar protections. One of these protections is a basic inactivity timeout, after which the session will be invalidated and reauthentication will be required by the application accessing the API. This is set to 30 seconds by default but can be disabled, thus leaving API sessions open indefinitely. The 30 second default must be verified and maintained.

The older an ESXi local account password is, the larger the opportunity window is for attackers to guess, crack or reuse a previously cracked password. Rotating passwords on a regular basis is a fundamental security practice and one that ESXi supports.

The CIM system provides an interface that enables hardware-level management from remote applications via a set of standard application programming interfaces (APIs). These APIs are consumed by external applications such as HP SIM or Dell OpenManage for agentless, remote hardware monitoring of the ESXi host. To reduce attack surface area and following the minimum functionality principal, the CIM service must be disabled unless explicitly needed and approved.

The default self-signed host certificate issued by the VMware Certificate Authority (VMCA) must be replaced with a DOD-approved certificate when the host will be accessed directly, such as during a virtual machine (VM) console connection. The use of a DOD certificate on the host assures clients the service they are connecting to is legitimate and properly secured.

While enabling Transmission Control Protocol (TCP) tunnels is a valuable function of sshd, this feature is not appropriate for use on the ESXi hypervisor.

OpenSLP implements the Service Location Protocol to help CIM clients discover CIM servers over TCP 427. This service is not widely needed and has had vulnerabilities exposed in the past. To reduce attack surface area and following the minimum functionality principal, the OpenSLP service must be disabled unless explicitly needed and approved. Note: Disabling the OpenSLP service may affect monitoring and third-party systems that use the WBEM DTMF protocols.

ESXi offers both local and remote audit recordkeeping to meet the requirements of the NIAP Virtualization Protection Profile and Server Virtualization Extended Package. Local records are stored on any accessible local or VMFS path. Remote records are sent to the global syslog servers configured elsewhere. To operate in the NIAP validated state, ESXi must enable and properly configure this audit system. This system is disabled by default. Note: Audit records can be viewed locally via the "/bin/viewAudit" utility over SSH or at the ESXi shell.

ESXi offers both local and remote audit recordkeeping to meet the requirements of the NIAP Virtualization Protection Profile and Server Virtualization Extended Package. Local records are stored on any accessible local or VMFS path. Remote records are sent to the global syslog servers configured elsewhere. To operate in the NIAP-validated state, ESXi must enable and properly configure this audit system. This system is disabled by default. Note: Audit records can be viewed locally via the "/bin/viewAudit" utility over SSH or at the ESXi shell.

When sending syslog data to a remote host via SSL, the ESXi host is presented with the endpoint's SSL server certificate. In addition to trust verification, configured elsewhere, this "x509-strict" option performs additional validity checks on CA root certificates during verification. These checks are generally not performed (CA roots are inherently trusted) and might cause incompatibilities with existing, misconfigured CA roots. The NIAP requirements in the Virtualization Protection Profile and Server Virtualization Extended Package, however, require even CA roots to pass validations.

Without establishing what types of events occurred, it would be difficult to establish, correlate, and investigate the events leading up to an outage or attack. Audit record content that may be necessary to satisfy this requirement includes, for example, time stamps, source and destination addresses, user/process/VM identifiers, event descriptions, success/fail indications, filenames involved, and access control or flow control rules invoked. Associating event types with detected events in the ESXi audit logs provides a means of investigating an attack; recognizing resource utilization or capacity thresholds; or identifying an improperly configured host.

Each VM on an ESXi host runs in its own "vmx" process. Upon creation, a vmx process will look in two locations for configuration items, the ESXi host itself and the per-vm *.vmx file in the VM storage path on the datastore. The settings on the ESXi host are read first and take precedence over settings in the *.vmx file. This can be a convenient way to set a setting in one place and have it apply to all VMs running on that host. The difficulty is in managing those settings and determining the effective state. Since managing per-VM vmx settings can be fully automated and customized while the ESXi setting cannot be easily queried, the ESXi configuration must not be used.

Each VM on an ESXi host runs in its own "vmx" process. Upon creation, a vmx process will look in two locations for configuration items, the ESXi host itself and the per-vm *.vmx file in the VM storage path on the datastore. The settings on the ESXi host are read first and take precedence over settings in the *.vmx file. This can be a convenient way to set a setting in one place and have it apply to all VMs running on that host. The difficulty is in managing those settings and determining the effective state. Since managing per-VM vmx settings can be fully automated and customized while the ESXi setting cannot be easily queried, the ESXi configuration must not be used.

An ESXi host's configuration consists of configuration files for each service that runs on the host. The configuration files typically reside in the /etc/ directory, but they can also reside in other namespaces. The configuration files contain run-time information about the state of the services. Over time, the default values in the configuration files might change, for example, when settings on the ESXi host are changed. A cron job backs up the ESXi configuration files periodically, when ESXi shuts down gracefully or on demand, and creates an archived configuration file in the boot bank. When ESXi reboots, it reads the archived configuration file and recreates the state that ESXi was in when the backup was taken. Before vSphere 7.0 Update 2, the archived ESXi configuration file is not encrypted. In vSphere 7.0 Update 2 and later, the archived configuration file is encrypted. When the ESXi host is configured with a Trusted Platform Module (TPM), the TPM is used to "seal" the configuration to the host, providing a strong security guarantee and additional protection from offline attacks. Configuration encryption uses the physical TPM when it is available and supported at install or upgrade time. If the TPM was added or enabled later, the ESXi host must be told to reconfigure to use the newly available TPM. Once the TPM configuration encryption is enabled, it cannot be disabled.

Unrestricted access to services running on an ESXi host can expose a host to outside attacks and unauthorized access. Reduce the risk by configuring the ESXi firewall to only allow access from authorized networks.

If a host is configured to join an Active Directory domain using Host Profiles and/or Auto Deploy, the Active Directory credentials are saved in the profile and are transmitted over the network. To avoid having to save Active Directory credentials in the Host Profile and to avoid transmitting Active Directory credentials over the network, use the vSphere Authentication Proxy.

When adding ESXi hosts to Active Directory, all user/group accounts assigned to the Active Directory group "ESX Admins" will have full administrative access to the host. If this group is not controlled or known to the system administrators, it may be used for inappropriate access to the host. Therefore, the default group must be changed to a site-specific Active Directory group and membership must be severely restricted.

ESXi can be configured to store log files on an in-memory file system. This occurs when the host's "/scratch" directory is linked to "/tmp/scratch". When this is done, only a single day's worth of logs are stored at any time. In addition, log files will be reinitialized upon each reboot. This presents a security risk as user activity logged on the host is only stored temporarily and will not persist across reboots. This can also complicate auditing and make it harder to monitor events and diagnose issues. ESXi host logging should always be configured to a persistent datastore. Note: Scratch space is configured automatically during installation or first boot of an ESXi host and does not usually need to be configured manually. If ESXi is installed on an SD card or USB device, a persistent log location may not be configured upon install as normal.

The "execInstalledOnly" advanced ESXi boot option, when set to TRUE, guarantees that the VMkernel executes only those binaries that have been packaged as part of a signed VIB. While this option is effective on its own, it can be further enhanced by telling the Secure Boot to check with the TPM to make sure that the boot process does not proceed unless this setting is enabled. This further protects against malicious offline changes to ESXi configuration to disable the "execInstalledOnly" option.

Starting in vSphere 8.0, the ESXi Entropy implementation supports the FIPS 140-3 and EAL4 certifications. Kernel boot options control which entropy sources to activate on an ESXi host. In computing, the term "entropy" refers to random characters and data that are collected for use in cryptography, such as generating encryption keys to secure data transmitted over a network. Entropy is required by security for generating keys and communicating securely over the network. Entropy is often collected from a variety of sources on a system. FIPS entropy handling is the default behavior if the following conditions are true: -The hardware supports RDSEED. -The disableHwrng VMkernel boot option is not present or is FALSE. -The entropySources VMkernel boot option is not present or is 0 (zero).

The log filtering capability allows users to modify the logging policy of the syslog service that is running on an ESXi host. Users can create log filters to reduce the number of repetitive entries in the ESXi logs and to deny specific log events entirely. Setting a limit to the amount of logging information restricts the ability to detect and respond to potential security issues or system failures properly.

Without confidentiality protection mechanisms, unauthorized individuals may gain access to sensitive information via a remote access session. In ESXi 8.0 Update 3 and newer TLS profiles are available to configure client and server TLS settings and must be configured to use only strong ciphers.

When using a standard key provider, the ESXi host relies on vCenter Server to manage the encryption keys. When using a trusted key provider, the ESXi host relies directly on the Trust Authority Hosts for keys, and vCenter Server is not involved. Regardless of the type of key provider, the ESXi host obtains the keys initially and retains them in its key cache. If the ESXi host reboots, it loses its key cache. The ESXi host then requests the keys again, either from the key server (standard key provider), or the Trust Authority Hosts (trusted key provider). When the ESXi host tries to obtain keys and the key server is offline or unreachable, vTPMs, vSAN encryption, and VM encryption cannot function. In vSphere, encrypted workloads can continue to function even when the key server is offline or unreachable. If the ESXi host has a TPM, the encryption keys are persisted in the TPM across reboots. So, even if an ESXi host reboots, the host does not need to request encryption keys. Also, encryption and decryption operations can continue when the key server is unavailable, because the keys have persisted in the TPM. If the encryption features of vSphere are used, it is to protect the confidentiality of workloads and while key persistence protects the availability of the environment it does so at the cost of confidentiality. An organization must consider the physical security posture and key provider reliability in their environments and if the risk of physical loss of equipment outweighs the availability of the workloads.

The dcui user is used for process isolation for the DCUI itself. The account has shell access which can be deactivated to reduce attack surface.

Hardware management controllers often present virtual or USB NICs to the ESXi host. These can be used as backdoors and should be deactivated both in the hardware configuration and in ESXi.