T0886: Remote Services

Require the authentication of devices and software processes where appropriate. Devices that connect remotely to other systems should require strong authentication to prevent spoofing of communications. Furthermore, software processes should also require authentication when accessing APIs.

The device or system should restrict read, manipulate, or execute privileges to only authenticated users who require access based on approved security policies. Role-based Access Control (RBAC) schemes can help reduce the overhead of assigning permissions to the large number of devices within an ICS. For example, IEC 62351 provides examples of roles used to support common system operations within the electric power sector (Citation: International Electrotechnical Commission July 2020), while IEEE 1686 defines standard permissions for users of IEDs. (Citation: Institute of Electrical and Electronics Engineers January 2014)

Architect sections of the network to isolate critical systems, functions, or resources. Use physical and logical segmentation to prevent access to potentially sensitive systems and information. Use a DMZ to contain any internet-facing services that should not be exposed from the internal network. Restrict network access to only required systems and services. In addition, prevent systems from other networks or business functions (e.g., enterprise) from accessing critical process control systems. For example, in IEC 62443, systems within the same secure level should be grouped into a zone, and access to that zone is restricted by a conduit, or mechanism to restrict data flows between zones by segmenting the network. (Citation: IEC February 2019) (Citation: IEC August 2013)

Set and enforce secure password policies for accounts.

Require user authentication before allowing access to data or accepting commands to a device. While strong multi-factor authentication is preferable, it is not always feasible within ICS environments. Performing strong user authentication also requires additional security controls and processes which are often the target of related adversarial techniques (e.g., Valid Accounts, Default Credentials). Therefore, associated ATT&CK mitigations should be considered in addition to this, including [Multi-factor Authentication](https://attack.mitre.org/mitigations/M0932), [Account Use Policies](https://attack.mitre.org/mitigations/M0936), [Password Policies](https://attack.mitre.org/mitigations/M0927), [User Account Management](https://attack.mitre.org/mitigations/M0918), [Privileged Account Management](https://attack.mitre.org/mitigations/M0926), and [User Account Control](https://attack.mitre.org/mitigations/M1052).

Network allowlists can be implemented through either host-based files or system hosts files to specify what connections (e.g., IP address, MAC address, port, protocol) can be made from a device. Allowlist techniques that operate at the application layer (e.g., DNP3, Modbus, HTTP) are addressed in [Filter Network Traffic](https://attack.mitre.org/mitigations/M0937) mitigation.

Manage the creation, modification, use, and permissions associated to user accounts.

Access Management technologies can be used to enforce authorization polices and decisions, especially when existing field devices do not provide sufficient capabilities to support user identification and authentication. (Citation: McCarthy, J et al. July 2018) These technologies typically utilize an in-line network device or gateway system to prevent access to unauthenticated users, while also integrating with an authentication service to first verify user credentials. (Citation: Centre for the Protection of National Infrastructure November 2010)

Use network appliances to filter ingress or egress traffic and perform protocol-based filtering. Configure software on endpoints to filter network traffic. Perform inline allow/denylisting of network messages based on the application layer (OSI Layer 7) protocol, especially for automation protocols. Application allowlists are beneficial when there are well-defined communication sequences, types, rates, or patterns needed during expected system operations. Application denylists may be needed if all acceptable communication sequences cannot be defined, but instead a set of known malicious uses can be denied (e.g., excessive communication attempts, shutdown messages, invalid commands). Devices performing these functions are often referred to as deep-packet inspection (DPI) firewalls, context-aware firewalls, or firewalls blocking specific automation/SCADA protocol aware firewalls. (Citation: Centre for the Protection of National Infrastructure February 2005)