Cisco Security
Contents
Introduction
Secure Operations
Monitor Cisco Security Advisories
Using Authentication, Authorization, and Accounting
Centralize Log Collection and Monitoring
Use of Secure Protocols
Traffic Visibility
Configuration Management
Management Plane
General Management Plane Hardening
Password Management
User Management
Administrative Model
Korn Shell and Auxiliary Authentication and Bypass
Disable Unused Services
Set Exec Timeout
Management Interfaces
Limit Network Access with Access Control Lists
Secured Interactive Management Sessions
Management Plane Protection
Control and Encrypt Management Sessions
Warning Banners
Authentication, Authorization, and Accounting
TACACS+ and RADIUS Authentication
Authentication Fallback
Redundant AAA Servers
Fortify Simple Network Management Protocol
SNMP Community Strings
SNMP Community Strings with Access Control Lists
Control SNMP Access with ACLs
Control SNMP with Management Plane Protection
SNMP Views
SNMP Version 3
Protect SNMP Private Community Strings
Logging Best Practices
AAA Logging
Access Control List Violation Logging
Logging Correlation
Send Logs to a Central Location
Logging Levels
Disable Console or Monitor Sessions
Buffered Logging
Configure Logging Source Interface
Configure Logging Timestamps
Configuration Management
Create Software and Configurations Backups
Exclusive Configuration Change Access
Control Plane
General Control Plane Hardening
IP ICMP Redirects
IP Direct-Broadcast
Disable IPv4 Source Routing
ICMP Destination Unreachable
IPv4 Options Packets
IPv4 Fragments Rate Limiting
IPv4 Fragments Filtering
IPv6 ICMP Rate Limiting
IPv6 Packet with Header Extension
TCP Service and Accept Rate Limiting
Proxy Address Resolution Protocol
Network Time Protocol
Limit CPU Impact of Control Plane Traffic
Control Plane Traffic
Local Packet Transport Services
CPU and Routing Protocol Software Queues
General Routing Protocol Securing Techniques
Message-Digest Algorithm 5 Peer Authentication
Keychain Management
Routing Policy Language
Secure Border Gateway Protocol
BGP Time-to-Live-Based Security Protection
BGP Peer Authentication with MD5
BGP Peer Authentication with Keychain
BGP Maximum Prefixes
Filter BGP Prefixes with RPL policies
Secure Interior Gateway Protocols
IGP TTL Security
IGP Peer Authentication with MD5
Open Shortest Path First Authentication with Keychain
IS-IS Authentication with Keychain
Passive Interface
IGP Route Filtering
IGP Routing Process Resource Consumption
Secure Label Distribution Protocol
Secure Resource Reservation Protocol
Secure First Hop Redundancy Protocols
Data Plane
Filter Transit Traffic with ACLs
IPv4 and IPv6 ACL with Counters
IPv6 ACL
ICMP Packet Filtering
Filter IPv4 Traffic with Remote Triggered Black Hole Filtering
Filter IPv6 Traffic with Remote Triggered Black Hole Filtering
Anti-Spoofing Protections
Unicast Reverse Path Forwarding
Anti-Spoofing ACLs
Limit CPU Impact of Data Plane Traffic
Features and Traffic Types that Impact the RP and LC CPU
Traffic Identification and Traceback
Identify Anomalous Activity by Using NetFlow
Identify Traffic by Using Classification ACLs
Conclusion
Glossary
Acknowledgments
References
Introduction
This document contains information that will help users secure Cisco IOS XR system devices to increase the overall security of a network. Structured around the three planes by which the functions of a network device are categorized, this document provides an overview of each Cisco IOS XR Software feature and references related documentation.
The three functional planes of a network, the management, control, and data planes, each provide a different functionality that must be protected.
Figure 1. Cisco IOS XR Software Architecture
The modular, physically, and logically distributed architecture of Cisco IOS XR Software, as shown in Figure 1, offers tremendous advantages in creating a highly available, secured routing platform and network. Discrete software components (subsystems) are implemented as separate software processes that run in their own protected memory address spaces. This implementation enables true fault isolation and compartmentalization in the event of a security incident by preventing faults in one subsystem from negatively affecting others.
The logical distribution of processes across three planes, each with its own access control for secure network operation, is the means to the deep fault isolation and implementation of security instrumentation within Cisco IOS XR Software.
Management Plane: The management plane contains the logical group of all traffic that supports provisioning, maintenance, and monitoring functions for the Cisco IOS XR device and the network. Traffic in this group includes SSH, FTP, Simple Network Management Protocol (SNMP), Syslog, TACACS+ and RADIUS, DNS, NetFlow, ROMMON, and Cisco Discovery Protocol. Management plane traffic is always destined to the local Cisco IOS XR device.
Control Plane: The control plane contains the logical group of all routing, signaling, link-state, and other control protocols that are used to create and maintain the state of the network and interfaces such as Border Gateway Protocol (BGP), Open Shortest Path First (OSPF), Label Distribution Protocol (LDP), Intermediate System to Intermediate System (IS-IS), and Address Resolution Protocol (ARP), and Layer 2 keepalive. Control plane traffic is always destined to the local Cisco IOS XR device.
Data Plane: The data plane contains the logical group of "customer" application traffic generated by hosts, clients, servers, and applications that are sourced from and destined to other similar devices supported by the network. Data plane traffic is mainly forwarded in the fast path and is never destined to the local Cisco IOS XR device.
The security features discussed in this document often provide enough detail for a network administrator (or operator) to configure the feature. However, in cases where it does not, the feature is explained to allow administrators to evaluate whether additional attention to the feature is required. Where possible and appropriate, this document contains recommendations that, if implemented, will help secure a network.
For more information about Cisco IOS devices, see the Cisco Guide to Harden Cisco IOS Devices.
Secure Operations
Secure network operations is a substantial topic. Although most of this document is devoted to the secure configuration of a Cisco IOS XR device, configurations alone do not completely secure a network. The operational procedures in use on the network, as well as the people who administer the network, contribute as much to security as the configuration of the underlying devices.
The following sections contain operational recommendations that network administrators are advised to implement. These sections highlight specific critical areas of network operations and are not comprehensive.
Monitor Cisco Security Advisories
The Cisco Product Security Incident Response Team (PSIRT) creates and maintains publications, commonly referred to as PSIRT Advisories, for security-related issues in Cisco products. Cisco PSIRT also publishes advisories to communicate less severe issues. Security advisories are available at http://www.cisco.com/go/psirt.
See the Cisco Security Vulnerability Policy for more information about Cisco PSIRT vulnerability reporting.
To maintain a secure network, network administrators should be aware of the information communicated in Cisco Security Advisories. Detailed knowledge of a vulnerability is required prior to evaluating the threat that the vulnerability can pose to a network. Refer to Risk Triage for Security Vulnerability Announcements for assistance with this evaluation process.
Using Authentication, Authorization, and Accounting
The Authentication, Authorization, and Accounting (AAA) framework is vital to securing network devices. The AAA framework provides authentication of management sessions, limits users to specific, administrator-defined commands, and logs all commands entered by all users.
See the Authentication, Authorization, and Accounting section of this guide for more information about leveraging AAA.
Centralize Log Collection and Monitoring
To gain an understanding of existing, emerging, and historic events that are related to security incidents, an organization should have a unified strategy for event logging and correlation. This strategy must leverage logging from all network devices and use pre packaged and customizable correlation capabilities.
After centralized logging is implemented, a structured approach must be developed to log analysis and incident tracking. Based on the needs of the organization, this approach can range from a simple diligent review of log data to an advanced rule-based analysis.
See the Logging Best Practices section of this guide for more information about how to implement logging on Cisco IOS XR network devices.
Use of Secure Protocols
Many protocols are used to carry sensitive network management data. Secure protocols should be used whenever possible. A secure protocol choice includes the use of SSH instead of Telnet so that both authentication data and management information are encrypted.
See the "Securing Interactive Management Sessions" section of this document for more information about the secure management of Cisco IOS XR devices.
Traffic Visibility
NetFlow provides the ability to monitor traffic flows in the network. Intended to optimally export traffic information to network management applications, NetFlow can also be used to show flow information on a router. This capability allows a network administrator to see what traffic traverses the network in real time. Even if flow information is exported to a remote collector, administrators are advised to configure network devices for NetFlow so the devices can be used reactively if necessary.
For more information about NetFlow, see the Cisco IOS NetFlow product information on Cisco.com.
Configuration Management
Configuration management is a process by which configuration changes are proposed, reviewed, approved, and deployed. In the context of a Cisco IOS XR device configuration, there are configure commit point records for each configuration change. These records can be used to determine what security changes were made and when these changes occurred. In conjunction with AAA log data, this information can assist in the security audit of network devices.
The configuration of a Cisco IOS XR device contains many sensitive details, including usernames, passwords, and the contents of access control lists (ACLs). The repository used to archive Cisco IOS XR device configurations should be secured and access should be restricted to only those roles and functions that require access. Insecure access to this information can undermine the security of the entire network.
Management Plane
The management plane consists of functions that achieve the management goals of the network. Such goals include interactive management sessions using SSH, as well as statistics gathering with SNMP or NetFlow. When considering the security of a network device, it is critical that the management plane is protected. If a security incident undermines the functions of the management plane, network recovery or stabilization may not be possible.
The following sections detail the security features and configurations available in Cisco IOS XR Software that help fortify the management plane.
General Management Plane Hardening
The management plane is used to access, configure, and manage a device, as well as monitor its operations and the network on which it is deployed. The management plane receives and sends traffic for operations of these functions. Both the management plane and control plane of a device must be secured, because operations of the control plane directly affect operations of the management plane. The following list includes protocols that are used by the management plane:
- Simple Network Management Protocol (SNMP)
- Telnet
- SSH
- SFTP
- FTP
- TFTP
- Secure Copy Protocol (SCP)
- TACACS+
- RADIUS
- NetFlow (also used by the Data Plane as that is where the traffic comes from)
- Network Time Protocol (NTP)
- Syslog
Administrators should take measures to ensure the survival of the management and control planes during security incidents. If one of these planes is successfully exploited, all planes can be compromised.
Password Management
Passwords control access to resources or devices, and administrators define passwords to authenticate requests. When a request is received for access to a resource or device, the request is challenged for verification of the password and identity, and access can be granted, denied, or limited based on the result. Security best practices dictate that passwords should be managed with a TACACS+ or RADIUS authentication server. However, a locally configured password for access is still required in the event that TACACS+ or RADIUS services fail. A device can also have other password information present within its configuration, such as network time protocol (NTP) key, a simple network management protocol community string, or routing protocol key, such as BGP and Message-Digest algorithm 5 (MD5) Authentication.
When a user configures a username and group membership in Cisco IOS XR Software, two types of passwords can be specified: one-way or two-way encrypted passwords. Secret passwords are ideal for user login accounts because the original, unencrypted password string cannot be deduced on the basis of the encrypted secret. Some applications, such as PPP, require only two-way passwords because they must decrypt the stored password for their own function, such as sending the password in a packet. For a login user, both types of passwords may be configured, but a warning message will display if one type of password is configured when the other is already present.
The following example shows the configuration for a Cisco IOS XR username, secret, and password. The secret is followed with a one-way password and the password is followed with a two-way password.
username user-a secret 5 $1$gAZC$gUwNDL./yo.aCSNpsG6lv1 password 7 13293E3C2E4C0723382727626771
If both the secret and password are configured for a user, the secret takes precedence for all operations that do not require a decryptable password, such as login. For applications like PPP, the two-way encrypted password is used even if a secret is present.
When possible, users should avoid logging in to Cisco IOS XR devices using an unencrypted protocol over any untrusted network. Users are advised to use encrypted login protocols, such as SSH or Kerberized Telnet; IPSec encryption for all management traffic, including Telnet, SNMP, and HTTP; or a one-time password system such as S/KEY or OPIE. Together with a TACACS+ or RADIUS server, encrypted login protocols provide control over interactive logins and privileged access.
User Management
Administrators should control access to resources or devices. When a request for access to a resource or device is received, the request is challenged for verification of the password and identity, and access can be granted, denied, or limited based on the result. Cisco IOS XR Software provides tools to allow for multiple levels of permissions using the concepts of task and user groups. User groups are defined to have access to a certain set of capabilities. Some of these capabilities are debug commands, show commands, and configuration commands. Different user groups have configuration access to different parts of the router. A root–system user must be created. The root-system user is the most powerful user in the Cisco IOS XR scheme and is essentially the same as a fully enabled (privilege level 15) user in Cisco IOS Software. The configuration for a root-system user follows:
username user_a password 7 1042081B group root-system
Cisco IOS XR Software allows the system administrator to configure groups of users and the job characteristics that are common to those groups. Groups must be explicitly assigned to users. Users are not assigned to groups by default. A user can be assigned to more than one group.
The following is a list of task groups that define different privileges for users in Cisco IOS XR Software. Each task group consists of the list of permitted tasks (read, write, execute, or debug permissions) for a user in the particular task group. Only root-system users have, by default, permission to run all tasks with the exception of the ones defined in the cisco-support group.
- cisco-support: Debugging and troubleshooting features, usually used by Cisco support personnel
- netadmin: Configuration tasks, such as those for routing protocols
- operator: Day to day monitoring activities and limited configuration rights
- root-system: Configuration and display rights for all secure domain routers in the system
- root-lr: Configuration and display rights for a specific secure domain router
- sysadmin: Administrative tasks such as maintaining the location for stored core dumps or setting up NTP.
- serviceadmin: Service administration tasks
A user group can derive attributes from another user group. Similarly, a task group can derive attributes from another task group.
Task Groups are defined by a collection of task IDs. Task groups contain task ID lists for each class of action. Each user group is associated with a set of task groups that are applicable to users in that group. A user’s task permissions are derived from the task groups that are associated with the user groups to which that user belongs.
The following example shows how to create a task group named mgmt that is inheriting attributes from the sysadmin task group. Read permissions are assigned for any CLI commands that are associated with task ID bgp.
taskgroup mgmt description backbone support functions inherit taskgroup sysadmin task read bgp
See the Configuring AAA Services on Cisco IOS XR Software section of the Cisco IOS XR System Security Configuration Guide for more information about configuring task groups.
Administrative Model
The router operates in two administrative planes: the administration (admin) plane and the secure domain router (SDR) plane. The admin (shared) plane consists of resources that are shared across all SDRs, while the SDR plane consists of resources that are specific to a particular SDR.
Secure Domain Router
For a single Cisco IOS XR device (CRS/GSR) chassis, several subscribers can connect to the chassis, each having different feature requests. Security best practices dictate dividing resources into different groups so that failures or security issues in one group do not impact other groups. The secure domain router (SDR) of Cisco IOS XR Software can help with this resource allocation.
Cisco routers running Cisco IOS XR Software can be partitioned into multiple, independent SDRs. The SDRs are a means of dividing a single physical system into multiple logically separated routers. The SDRs perform routing functions identical to a physical router, but they share resources with the rest of the system. For example, the software, configurations, protocols, and routing tables that are assigned to an SDR belong to that SDR only, but other functions, such as chassis control and switch fabric, are shared with the rest of the system. Because the CPU and memory of an SDR are not shared with other SDRs in the chassis, the failures, security attacks, and configuration problems that cause out-of-resources conditions in one SDR do not affect other SDRs.
The following example shows how to create an SDR on a Cisco IOS XR 12000 Series Router. A single SDR that is named as SDR-1 is created and has three nodes (0/0/* 0/1/* 0/5/*):
admin configure sdr SDR-1 location 0/0/* location 0/1/* location 0/5/* end
The root-system user has the highest level of responsibility for the router. The provisions of this user-secure domain routes and create root SDR users. The root SDR users take most of the responsibilities from the root-system user for the SDR. The root SDR users in turn can create SDR users. The root-system users and root SDR users have fixed permissions (task IDs) that cannot be changed by other users.
Each SDR has its own AAA configuration that includes local users, groups, and TACACS+ and RADIUS configurations. Users that are created in one SDR cannot access other SDRs unless the same users are configured in those other SDRs.
Administrative Access
Administrative access to the system can be lost if administrators do not fully understand or carefully plan for the following operations. A lockout of all root-system users is a serious issue that requires a system reload to recover the password.
- Configuring authentication that uses remote AAA servers that are not available, particularly authentication for the console.
- Removing the flash card from disk0:, or a disk corruption may deny auxiliary port authentication, can affect certain system debugging abilities. However, if the console is available, the system is still accessible.
- Configuring command authorization or EXEC authorization on the console should be done with extreme care, because TACACS+ servers may not be available or may deny every command, thus locking the user out. In particular, this lockout can occur if the authentication takes place with an unknown TACACS+ server user, or if the TACACS+ user has most or all the commands denied.
To avoid a lockout, administrators are advised to perform one of the following actions:
- Before turning on TACACS+ command authorization or EXEC authorization on the console, ensure that the user who is configuring the authorization is logged in with the appropriate user permissions in the TACACS+ profile.
- If the security policy of the site permits, use the none option for command authorization or EXEC authorization. In the event TACACS+ servers are not accessible, this options enables AAA to roll over to the none method, permitting the user to run the command.
- In the event of failure of the TACACS+ or RADIUS services, locally configure passwords for access.
Korn Shell and Auxiliary Authentication and Bypass
The korn shell (Ksh) is the primary shell for the auxiliary port of route processor (RP), standby RP, and distributed RP cards and for console and auxiliary ports of line cards (LCs) and service processors (SPs). The following are some of the characteristics of Ksh authentication:
- For security purposes, Ksh/Aux authentication allows only root-system users who have a secret configured.
- For Cisco IOS XR Software versions prior to 3.8, Ksh/Aux does not authenticate TACACS+ or RADIUS users, even if they are root-system users. For Cisco IOS XR Software versions 3.8 and later and Cisco IOS XR Software version 3.6.2, Ksh/Aux supports TACACS+/RADIUS authentication.
- Because Ksh/Aux authentication uses a single user password database, when a root-system user on a dSC (Designated Shelf Controller) is configured using the normal AAA CLI, that user can log in using the Ksh/Aux authentication username and password for any card. This includes the RP, standby RP, LC, and SP.
- Ksh/Aux authentication cannot be turned off or bypassed after the card is booted. To bypass authentication, a user needs to reload the card. For more details, see the Korn Shell and Auxiliary Authentication and Bypass section of this guide.
There are cases when Ksh/Aux authentication must be bypassed, including the following scenarios:
- Designated system controller (dSC) ACTIVE RP disk0 corruption
- Loss of Qnet connectivity
- Inability to determine the node ID of the dSC ACTIVE RP
To bypass Ksh/Aux authentication, the user must set the ROMMON variable AUX_AUTHEN_LEVEL to 0 and then reload the image. A reboot is required only on the card that must bypass authentication.
The ROMMON variable AUX_AUTHEN_LEVEL can have one of the following values:
- 0–Authentication will be bypassed on the card.
- 1–Loose authentication. Authentication is performed on a best-effort basis and permits the user to access Ksh/Aux if the system cannot access authentication information successfully.
- 2–Strict authentication. This is the default state.
To bypass authentication on the card, issue the following commands under rommon:
rommon1> AUX_AUTHEN_LEVEL=0 rommon2> sync rommon2> boot tftp:/ ...
Note: Administrators are advised to provide Auxiliary (Aux) port access only to dSC root-system users. SDR users should not have physical Aux port access because they can log in to the system with root privileges and obtain access to any router resource including the ones outside the SDR; this type of access is not desirable.
Disable Unused Services
As a security best practice, administrators should disable unnecessary services. Most services are disabled by default in Cisco IOS XR Software; however, these services can be enabled by issuing their respective configuration commands. Administrators are advised to disable the following services if they are not necessary for business operations
- TCP and UDP small services:
- echo (port number 7)
- discard (port number 9)
- daytime (port number 13)
- chargen (port number 19)
- HTTP
- Cisco Discovery Protocol (DP)
- Simple Network Management Protocol (SNMP)
- TFTP
- DHCP
To disable TCP and UDP small services, issue the following command:
no service {ipv4|ipv6} tcp-small-servers
no service {ipv4|ipv6} udp-small-servers To disable HTTP services, issue the following command:
no http server
Cisco Discovery Protocol (DP) can be disabled globally or under interface. To disable Cisco DP services, issue the following command:
no cdp interface pos 0/0/0/1 no cdp
To disable Simple Network Management Protocol (SNMP) services, issue the following command:
no snmp-server
To disable TFTP services, issue the following command:
no tftp ipv4 server no tftp ipv6 server
To disable DHCP services if DHCP relay services are not required, issue the following command:
no dhcp {ipv4|ipv6}If a service is required, administrators can leverage the MPP feature to enable and disable a service on the specified interface. See the Management Plane Protection section of this guide for more information.
Set Exec Timeout
By default, console, vty, and tty sessions disconnect after 10 minutes of inactivity. Administrators are advised to maintain this value at 10 minutes or less but greater than zero. A 0-minute value will prevent sessions from terminating.
Issue the following command to set an exec-timeout value:
line {console|default|template} exec-timeout x y Issue the following command to reinstate the default timeout value:
no line {console|default|template} exec-timeoutManagement Interfaces
The management plane of a Cisco IOS XR device is accessed in-band or out-of-band on a physical or logical management interface. Ideally, both in-band and out-of-band management access exists for each network device so that the management plane can be accessed during network outages.
One of the most common interfaces used for in-band access is the logical loopback interface. Loopback interfaces are always available, whereas physical interfaces can change state, resulting in the interface and subsequently the device not being accessible. Administrators are advised to add a loopback interface to each device as a management interface and to use it exclusively for the management plane.
In addition to loopback interfaces, administrators are advised to use the Virtual Address feature in combination with uniquely addressed management interfaces:
ipv4 virtual address 12.7.49.171 255.255.0.0 "Virtual (shared) Address"
!
interface MgmtEth0/0/CPU0/0 "Primary PRP"
ipv4 address 12.7.49.172 255.255.0.0
!
interface MgmtEth0/1/CPU0/0 "Secondary PRP"
ipv4 address 12.7.49.173 255.255.0.0
!
To ensure that UDP traffic flow, such as syslog and SNMP traps, is not interrupted during the failover, administrators are advised to use virtual addresses to source all management traffic, as shown in the following configuration:
ipv4 virtual address use-as-src-addr
This feature is part of Cisco IOS XR Release 3.6.2 and later.
Limit Network Access with Access Control Lists
Devised to prevent unauthorized direct communication to network devices, infrastructure ACLs are one of the most critical security control mechanisms that can be implemented in the network.
An ACL is constructed and applied to specify necessary connections between hosts or networks and network devices. Common examples of these types of connections are eBGP, SSH, and SNMP. After the required connections have been permitted, all other traffic to the infrastructure is explicitly denied. All transit traffic that crosses the network and is not destined to infrastructure devices is explicitly permitted.
The protection provided by ACLs is relevant to both the management and control planes. The implementation of ACLs can be made easier through the use of distinct addressing for network infrastructure devices.
The following ACL configuration example illustrates the structure that is required as a basis for starting the ACL implementation process:
ipv4 access-list ACL-INFRASTRUCTURE-IN !—-Permit required connections for routing protocols and 10 permit tcp host <trusted-ebgp-peer> host <local-ebgp-address> eq 179 20 permit tcp host <trusted-ebgp-peer> eq 179 host <local-ebgp-address> 30 permit tcp host <trusted-management-stations> any eq 22 40 permit udp host <trusted-netmgmt-servers> any eq 161 !—Deny all other IP traffic to any network device ! 50 Deny ipv4 any <infrastructure-address-space> <mask> !—Permit transit traffic ! 60 permit ipv4 any any
When created, the ACL must be applied to all interfaces that face non-infrastructure devices, including interfaces that connect to other organizations, remote access segments, user segments, and segments in data centers.
Secured Interactive Management Sessions
Management sessions allow administrators to view and collect information about a Cisco IOS XR device and its operations. If this information is disclosed to a malicious user, the device can become the target of an attack or used as a source of additional attacks. Anyone with privileged access to a Cisco IOS XR device has the capability for full administrative control of the device. It is imperative to secure management sessions to prevent information disclosure and unauthorized access.
Management Plane Protection
The Management Plane Protection (MPP) feature in Cisco IOS XR Software provides the capability to restrict the interfaces on which network management packets are allowed to enter a device. MPP allows a network operator to reserve a set of interfaces for management traffic either exclusively, or along with regular data. MPP changes the default behavior of an interface by not listening to all the services traffic and allowing traffic only from applications that are explicitly configured on the interface.
MPP provides the support for in-band management interfaces. An in-band interface shares management and forwarding traffic.
MPP provides a mechanism for specifying configurable out-of-band interfaces. An out-of-band management interface receives network management traffic. The advantage of this implementation is that network management traffic does not interfere with any forwarding or customer traffic, significantly reducing the possibility of potential side effect, such as garbled packet and a denial of service (DoS) condition, of processing network management traffic. MPP also supports peer-filtering mechanisms. Management traffic that belongs to the configured peer ip-addresses would be allowed on a specified interface; other traffic would be dropped.
In-band management interface: This Cisco IOS XR physical or logical interface processes management packets, as well as data-forwarding packets. An in-band management interface is also called a shared management interface.
Out-of-band interface: This interface allows only management protocol traffic to be forwarded or processed. An out-of-band management interface is defined by the network operator to specifically receive network management traffic. The advantage is that forwarding (or customer) traffic cannot interfere with the management of the router, which significantly reduces the possibility of DoS attacks.
Out-of-band interfaces: This interface forwards traffic only between out-of-band interfaces or terminate management packets that are destined to the router. In addition, the out-of-band interfaces can participate in dynamic routing protocols. The service provider connects to the router’s out-of-band interfaces and builds an independent overlay management network, with all the routing and policy tools that the router can provide. Management ports are out-of-band by default.
Device management traffic may enter a device only through these management interfaces. After MPP is enabled, no interfaces except the designated management interfaces will accept network management traffic that is destined to the device. Restricting management packets to the designated interfaces provides greater control over the management of a device, providing more security for that device. Administrators are advised to use the peer-filtering option to control management traffic from specific peers or a range of peers.
MPP configuration does not enable the specific management protocol services. MPP is responsible only for making the services available on different interfaces. Currently, MPP only controls the incoming management requests for protocols, such as TFTP, Telnet, Simple Network Management Protocol (SNMP), Secure Shell (SSH), and HTTP and HTTPS.
The following example shows how to enable the MPP to only allow SSH and HTTPS requests from peer address 10.0.1.0 on the GigabitEthernet0/0/01 interface:
control-plane management-plane inband interface Gig 0/0/01 allow ssh peer address ipv4 10.0.1.0/24 allow https peer address ipv4 10.0.1.0/24
The following example shows how to configure GigabitEthernet0/0/0/1 as an out-of-band interface and only allow a Telnet request from peer address 10.10.10.10:
control-plane management-plane out-of-band interface GigabitEthernet0/2/0/0 allow Telnet peer address ipv4 10.10.10.10
Note: In Cisco IOS XR version 3.8.0 and later, MPP is enabled on all interfaces by default. Any TCP and UDP port can be accessed from any interface. However, when MPP is configured, access is only permitted on management interfaces until the MPP permit is applied to other interfaces.
For more information on management plane protection, see the Implementing Management Plane Protection on Cisco IOS XR Software section in the Cisco IOS XR System Security Configuration Guide.
Control and Encrypt Management Sessions
Because information can be disclosed during an interactive management session, traffic must be encrypted so that a malicious user cannot access the data that is being transmitted. Encrypting the traffic allows a secure remote access connection to the device. If the traffic for a management session is sent over the network in plain text, an attacker could obtain sensitive information about the device and the network.
Telnet Protocol
Although popular, Telnet is not a secure protocol, and administrators of Cisco IOS XR devices are advised not to use Telnet. The use of Telnet on Cisco IOS XR devices will require special attention.
The Cisco Internet services process daemon, Cinetd, which is similar to the UNIX daemon, inetd, is a multi-threaded server process that is started by the system manager after the system has booted. Cinetd listens on a well-known port on behalf of the server program. When a service request is received on the particular port, Cinetd notifies the server program that is associated with the service request. By default, Cinetd is not configured to listen for any services.
Telnet service is disabled by default on Cisco IOS XR. The following command is required to enable Telnet service on a Cisco IOS XR device:
telnet [ipv4 | ipv6 | vrf vrf-name] server max-servers [option] <1-200> Set number of allowable Telnet servers no-limit No limit to number of allowable Telnet servers
Two options are available to define the maximum allowable Telnet servers and to set "no limit" on the number of allowable Telnet servers.
Note: Administrators are advised not to choose the no-limit option, because the option could bring the system to an unstable state. Cisco IOS XR Software versions 3.8 and later do not include the no-limit option.
To add protection using Telnet service, the following configuration is recommended:
- Restrict Telnet access using ACL, MPP, and peer control.
- Estimate the maximum Telnet sessions that the system may need to set Telnet max-servers accordingly.
- Add a hardware rate-limit for Telnet packets using the Local Packet Transport Services (LPTS) feature, if necessary. The default value of TELNET-known is 1000 packets per second (p/s). The default value of TELNET-default is 500 p/s. The default value can be changed to a lower value as shown in the following example:
# below is global configure across all LCs lpts pifib hardware police flow telnet default rate 200 flow telnet known rate 200 # below is specific LC configure which has precedence than global configure lpts pifib hardware police location 0/1/CPU0 flow telnet default rate 50 flow telnet known rate 200
Note: The 200 p/s is a sample number for reference only.
This feature is available in Cisco IOS XR Software versions 3.6 and later.
For more information about securing Telnet protocols, see the Cisco IOS XR IP Addresses and Services Configuration Guide and Cisco IOS XR IP Addresses and Services Command Reference.
SSH Protocol
Users can establish an encrypted and secure remote access management connection to a device by using the SSH, SFTP, or HTTPS protocols. Cisco IOS XR Software supports SSH Version 1.0 (SSHv1), SSH Version 2.0 (SSHv2), and HTTPS that uses SSL and Transport Layer Security (TLS) for authentication and data encryption.
The following example configuration enables SSHv1 or SSHv2 on a Cisco IOS XR device:
hostname test-1 domain name test.com ssh server (or ssh server v2)
Cisco IOS XR Software also supports SFTP, which provides an encrypted, secure, and authenticated connection for copying device configurations or software images. SFTP relies on SSHv2. After SSHv2 has been configured, the SFTP feature is available on the router.
Both versions of SSH require more RP CPU resources than Telnet that could cause higher RP CPU usage during high-rate SSH requests or an SSH DoS attack. The following configuration can help prevent excessive RP CPU usage:
- Restrict the SSH access using MPP and peer control
- In Cisco IOS XR, the default rate-limit of the SSH server is 60 requests per minute; users can change this rate to a lower value.
ssh server rate-limit 10
- Rate-limit the SSH packets in the hardware layer using LPTS. The default value is 1000 p/s for Cisco IOS XR Software versions 3.6 and later. This value can be changed to a lower value, as shown in the following example:
# below is global configure across all LCs
lpts pifib hardware police
flow ssh default rate 200
flow ssh known rate 200
# below is is specific LC configure which has precedence than global configure
lpts pifib hardware police location 0/1/CPU0
flow ssh default rate 50
flow ssh known rate 200HTTP and HTTPS
Cisco IOS XR Software may use the craft web interface (CWI) to support remote configuration and monitoring. In general, HTTP access is equivalent to interactive access to the router. The authentication protocol used for HTTP is equivalent to sending a plaintext password across the network and there is no effective provision in HTTP for challenge-based or one-time passwords. HTTP access should be restricted to appropriate IP addresses that use the http server access-group command or MPP with peer control. As with interactive logins, the best choice for HTTP authentication is a TACACS+ or RADIUS server. Administrators are advised to avoid using "enable" as the password for the HTTP authentication.
HTTPS (HTTP over SSL or HTTP Secure) is the use of SSL or Transport Layer Security (TLS) as a sub-layer under regular HTTP application layering. HTTPS encrypts and decrypts user page requests as well as the pages that are returned by the web server. Because HTTPS protects against eavesdropping and man-in-the-middle attacks, administrators are advised to use HTTPS instead of HTTP.
HTTPS can be enabled only after SSL initialization. To initialize SSL, Rivest, Shamir, and Adelman (RSA) or Digital Signature Algorithm (DSA), the following actions need to be performed:
- The key pairs need to be generated
- A Certificate Authority (CA) needs to be enrolled
- A CA certificate for the router key needs to be obtained
The following example shows how to generate the RSA keys for the router, configure a trust point, authenticate to the CA server, and obtain a key certificate from the CA:
crypto key generate rsa general-keys configure domain ipv4 host xyz-ultra5 10.0.0.5 crypto ca trustpoint myca enrollment url http://xyz-ultra ! crypto ca authenticate myca crypto ca enroll myca
The following example shows how to enable HTTPS:
http server ssl
Control Physical and Virtual Terminals
In Cisco IOS XR devices, physical and virtual terminals are "lines" that can be used for local and remote access to a device.
Note: The console ports on Cisco IOS XR devices have special privileges that allow an administrator to perform the password recovery procedure. An unauthenticated attacker who is trying to perform password recovery will require access to the console port and be able to interrupt power to the device or cause a crash of the device.
Any method that is used to access the console port of a device must be secured in a manner that is equal to the security that is enforced for privileged access to a device. Methods used to secure access must include the use of AAA, exec-timeout, and modem passwords (if a modem is attached to the console).
In most situations, the AUX port of a device should be accessible only by a root-system user.
The physical terminal line for the console port is identified by its location, which is expressed in the format of rack/shelf/module on the active or standby RP where the console and AUX port reside.
Physical location is not applicable for virtual terminals. The Cisco IOS XR Software assigns a vty identifier to the vty lines and connections according to the order in which the vty connection was established. A vty line is used for remote network connections that are supported by the device regardless of protocol (for example, SSH, SCP, or Telnet). To ensure that a device can be accessed by way of a local or remote management session, proper controls must be enforced on the vty lines. Cisco IOS XR devices have a limited number of vty lines; the number of lines available can be determined by issuing the show line EXEC command. When all the vty lines are in use, new management sessions cannot be established, resulting in a DoS condition on the device.
The terminals are configured by using line templates. The following line templates are available in the Cisco IOS XR Software:
- Default line template: This template applies to all physical and virtual terminal lines.
- Console line template: This template applies to the console line.
- Line templates: This is a user-defined line template that can be applied to a range of virtual terminal lines.
Attributes that are not defined in the console template or in any virtual template are taken from the default template. Administrators can secure the console by using the following configuration:
line console exec-timeot 60 0 secret s3cr3t
The simplest form of access control to the vty of a device is through the use of authentication on all lines, regardless of the device location within the network. This authentication is critical for vty lines because they are accessible by way of the network. Other forms of vty access controls can be enforced by using the transport input or access-class configuration commands, using the MPP and LPTS features, or by applying ACLs to the physical interfaces on the device.
Authentication can be enforced through the use of AAA through the local user database or by simple password authentication configured directly on the vty line. AAA is the recommended method for authenticated access to a device.
The exec-timeout command must be used to log out sessions on vty lines that are left idle.
A vty should be configured to accept only encrypted and secure remote access management connections to the device using the following configuration:
line {default | console | temple}
transport input ssh
vty lines allow an administrator to connect to other devices. Administrators are advised to use the transport output line configuration command to limit the type of transport outgoing connections. If outgoing connections are not necessary, then transport output none should be used. However, if outgoing connections are allowed, an encrypted and secure remote access method for the connection should be enforced through the use of transport output SSH.
The following is an example that shows a configuration for securing vty lines:
line default
access-class 20 in
exec-timeout 60 0
password s3cr3t
transport preferred ssh
Any vty should be configured to accept connections with only the protocols that are necessary. Configuration can be done using the transport input command. For example, a vty that is expected to receive only Telnet sessions would be configured with transport input Telnet, while a vty that permits both Telnet and SSH sessions would have transport input Telnet SSH. If the software supports an encrypted access protocol, such as SSH, administrators are advised to enable only that protocol and to disable plaintext Telnet. Administrators also may consider using the ip access-class command to restrict the IP addresses from which the vty will accept connections. Again, MPP and LPTS are the recommended solutions for limiting packet rate to the system at a hardware level. ACLs should be used as a secondary protection in case MPP and LPTS are not configured properly.
Control vtys and Ensure vty Availability
Cisco IOS XR supports up to 100 vty lines (its default number of vty lines is five), and they are managed by configuring vty-pool template (vty-pool default 0 99). When all of the vtys are in use, further remote interactive connections cannot be established, creating the opportunity for a DoS attack; if an attacker can open remote sessions to all the system vtys, the legitimate administrator may not be able to log in. An exploit does not require the attacker to log in using a valid username and password because the sessions can be left at the login prompt.
Configuring a more restrictive ip access-class command on the last system vty may reduce the likelihood of an attack. The last vty could be restricted to accept connections from a single, specific administrative workstation, whereas the other vtys may accept connections from any address in a corporate network.
The following configuration example shows that vty ports 5 and 6 accept connections only from IP source address 10.10.10.10 (for example, administrator host), while ports 0–4 can accept connections from any source.
ipv4 access-list vty5-permit
10 permit ipv4 host 10.10.10.10 any
20 deny ipv4 any any
!
line template vty5
password 7 0822455D0A16544541
access-class ingress vty5-permit
!
vty-pool vty5 5 6 line-template vty5
vty-pool default 0 4Cisco IOS XR devices have default vty timeouts of 30 seconds if there is no response on a login prompt. These timeouts prevent idle sessions from consuming a vty indefinitely and also have default timeout if any open vty sessions are idle for more than 5 minutes. Because there is no TCP keep alive command in Cisco IOS XR Software, these timeouts help guard against both malicious attacks and "orphaned" sessions that are caused by remote system crashes.
Complete vty protection can be provided by disabling all non-IP based remote access protocols and using IPsec encryption for all remote interactive connections to the router.
Warning Banners
In some jurisdictions, it may be impossible to prosecute malicious users and illegal to monitor them unless they have been notified that they are not permitted to use the system. One method of notification is to place this information into a banner message that is configured using the Cisco IOS XR Software banner login command.
Legal notification requirements are complex, vary by jurisdiction and situation, and should be discussed with legal counsel. Legal opinions can vary within the same jurisdiction. In cooperation with legal counsel, a banner should provide some or all of the following information:
- Notice that the system is to be logged into or used only by specifically authorized personnel with information about who can authorize use
- Notice that unauthorized use of the system is unlawful and may be subject to civil and criminal penalties
- Notice that use of the system can be logged or monitored without further notice and that the resulting logs can be used as evidence in court
- Specific notices required by local laws
From a security point of view, a login banner should not contain any specific information about the router name, model, software, or ownership. This information can be exploited by malicious users. The following is an example of the banner configuration where "c" is a delimiting character:
banner motd c This is restricted to access c
Authentication, Authorization, and Accounting
The Authentication, Authorization, and Accounting (AAA) framework is critical to securing interactive access to network devices. The AAA framework provides a highly configurable environment that can be tailored depending on the needs of the network.
Authentication is the most important security process by which a principal (a user or an application) obtains access to the system. The principal is identified by a username or user ID that is unique across an administrative domain. The applications serving the user, such as EXEC or Management Agent, obtain the username and the credentials from the user. AAA performs the authentication based on the username and credentials that are passed to it by the applications. The role of an authenticated user is determined by the group (or groups) to which the user belongs. A user can be a member of one or more user groups.
The root-system user can log in to any node in any SDR in the system. A user is a root-system user if they belong to the root-system group. The root-system user may be defined in the local or remote AAA database.
When logging in from a non-owner SDR, the root-system user must add the "@admin" suffix to the username. The use of the "@admin" suffix will send the login request authentication to the SDR owner/dSC (designated shelf controller) for verification. The dSC uses an authentication method that is configured by the aaa authentication login remote command to accept or reject the login request.
AAA can be enabled on a Cisco IOS XR device using a configuration similar to the following example:
aaa authentication login default group tacacs+ local
The following example shows how to enable the Remote method using the admin configure mode:
aaa authentication login remote group tacacs+ local
See the Configuring AAA Services on Cisco IOS XR Software section of the Cisco IOS XR System Security Configuration Guide for more information about AAA.
TACACS+ and RADIUS Authentication
Cisco IOS XR Software communicates with the remote AAA server using a standard IP-based security protocol, such as TACACS+ or RADIUS.
TACACS+ uses a TCP port (49), while RADIUS uses UDP. TACACS+ provides reliable data transfer between the client and server; RADIUS provides faster operation. Benefits to using TACACS+ include the following:
- TACACS+ allows true command authorization. Users can create clear usage policies with TACACS+, whereas different users have access to different commands with administrative granularity. TACACS+ can do this because it separates the Authentication and Authorization functions. RADIUS combines these functions.
- TACACS+ encrypts the entire payload of the client-server exchange. This action is important for the protection of highly secure environments. In contrast, RADIUS encrypts only the password, allowing intercepted packets to reveal important information.
The strongest point in favor of RADIUS is that it is an open standard that is implemented by many vendors, including Cisco.
For more information about the differences between RADIUS and TACACS+, see TACACS+ and RADIUS Comparison.
The following sample configuration shows how TACACS+ authentication can be enabled on a Cisco IOS XR device:
tacacs-server host 10.1.1.1 port 49 key cisco123 tacacs source-interface Loopback0
The following sample configuration shows how RADIUS authentication can be enabled on a Cisco IOS XR device:
radius-server host 5.5.151.1 auth-port 1812 acct-port 1813 key cisco123 !
Authentication Fallback
If all configured TACACS+ servers become unavailable, a Cisco IOS XR device can rely on secondary authentication protocols. Typical configurations include the use of local database authentication if all configured TACACS+ servers are unavailable.
The complete list of options for on-device authentication includes local or line. Line authentication uses passwords that are configured under line device statements such as console or default templates, while local authentication sets passwords that can be used on all lines. Both options have advantages. There are two options for password configuration: the use of clear text via the password command, and the use of encrypted secure password by configuring the secret option. The secret option is preferred.
If TACACS+ became completely unavailable, each administrator could use their local username and password. Although this action enhances the accountability of network administrators during TACACS+ outages, it significantly increases the administrative burden, because local user accounts on all network devices must be configured and subsequently maintained.
The following configuration example builds on the previous TACACS+ authentication example to include fallback authentication to the locally configured password, or to the line password if no local password is configured:
aaa authentication login default group tacacs+ local line
Redundant AAA Servers
Whether TACACS+ or RADIUS servers are used, the AAA servers should be redundant and deployed in a fault-tolerant manner. These attributes can help ensure that interactive management access such as SSH is possible if an AAA server is not available. Administrators are advised to consider the following factors when designing AAA servers:
- Availability of AAA servers during potential network failures
- Geographically dispersed placement of AAA servers
- Load on individual AAA servers during steady-state and failure conditions
- Network latency between Network Access Servers and AAA servers
- AAA server databases synchronization
The following example shows how to create redundancy by configuring a total of three TACACS+ servers to be used by the Cisco IOS XR device:
tacacs-server host 1.1.1.1 port 1 key abc tacacs-server host 2.2.2.2 port 2 key def tacacs-server host 3.3.3.3 port 3 key ghi aaa group server tacacs+ tacgrp server 1.1.1.1 server 2.2.2.2 server 3.3.3.3
Fortify Simple Network Management Protocol
This section highlights methods that can be used to secure the deployment of Simple Network Management Protocol (SNMP) within Cisco IOS XR devices. SNMP must be properly secured to protect the confidentiality, integrity, and availability of the network data and the network devices through which the data transits. SNMP provides a wealth of information on the health of network devices. This information should be protected from malicious users who may want to take advantage of the data for use in attacks against the network.
SNMP Community Strings
Community strings are passwords that are applied to a Cisco IOS XR device to restrict access, both read-only and read-write, to the SNMP data on the device. Community strings, as with all passwords, should be carefully chosen to conform to security best practices. Community strings should be changed at regular intervals and in accordance with network security policies. For example, the strings should be changed when a network administrator changes roles or leaves the company.
The following example shows a sample configuration for a global SNMP community string, s3cr3t:
snmp-server community s3cr3t
The following configuration lines illustrate the configuration of a read-only community string of r3adm3 and a read-write community string of s3cr3t:
snmp-server community r3adm3 RO snmp-server community s3cr3t RW
Note: The preceding community string examples have been chosen to clearly explain the use of community strings. For production environments, community strings should be chosen with caution and should consist of a series of alphabetical, numerical, and non-alphanumeric symbols.
To limit SDR SNMP access, the SDRowner or SystemOwner options can be used in the following manner:
snmp-server community r3adm3 RO SDRowner snmp-server community r3adm3 RO SystemOwner
SNMPv2 Community strings are displayed in plain text. Administrators are advised to change community strings on a periodic basis as changing passwords.
SNMP Community Strings with Access Control Lists
In addition to the community string, an ACL should be applied that further restricts SNMP access to a select group of source IP addresses. The following configuration example restricts SNMP read-only access to end host devices that reside in the 192.168.100.0/24 address space and restricts SNMP read-write access to only the end host device at 192.168.100.1.
Note: The devices that are permitted by these ACLs still require the proper community string to access the requested SNMP information.
ipv4 access-list SNMP-ROACCESS 10 permit ipv4 192.168.100.0 0.0.0.255 any ipv4 access-list SNMP-RWACCESS 10 permit ipv4 host 192.168.100.1 any snmp-server community r3adm3 RO SNMP-ROACCESS snmp-server community s3cr3t RW SNMP-RWACCESS
Control SNMP Access with ACLs
Infrastructure ACLs can be deployed to ensure that only end hosts with trusted IP addresses can send SNMP traffic to a Cisco IOS XR device. An ACL should contain a policy that denies unauthorized SNMP packets on UDP port 161.
Control SNMP with Management Plane Protection
The Management Plane Protection (MPP) feature in Cisco IOS XR Software can be used to help secure SNMP by restricting the interfaces through which SNMP traffic can terminate on the device. The following example shows POS 0/11/0/0 as the interface that will terminate SNMP traffic from host 1.1.1.1:
control-plane management-plane inband interface pos 0/11/0/0 allow snmp peer address ipv4 1.1.1.1/32
SNMP Views
SNMP Views is a security feature that can permit or deny access to certain SNMP MIBs. When a view is created and applied to a community string with the snmp-server community community-string view global configuration commands, access to MIB data is restricted by the permissions that are defined with the view. When appropriate, users are advised to use views to limit access to specific data to SNMP users.
The following configuration example restricts SNMP access with the community string LIMITED to the MIB data that is located in the VIEW−SYSTEM−ONLY group, as configured in the first three lines of the following example using the snmp-server view command:
snmp-server view VIEW-SYSTEM-ONLY ciscoPingMIB included snmp-server view VIEW-SYSTEM-ONLY sysUpTime included snmp-server view VIEW-SYSTEM-ONLY system include snmp-server community LIMITED view VIEW-SYSTEM-ONLY RO
SNMP Version 3
SNMP Version 3 (SNMPv3) is defined by RFCs 3410 through 3415 and is an interoperable, standards-based protocol for network management. SNMPv3 provides secure access to devices by authenticating and optionally encrypting packets over the network. Where supported, SNMPv3 can be used to add another layer of security when deploying SNMP. SNMPv3 consists of three primary configuration options:
- no auth: This mode does not require authentication or encryption of SNMP packets.
- auth: This mode requires authentication of the SNMP packets without encryption.
- priv: This mode requires both authentication and encryption (privacy) of each SNMP packet.
An authoritative engine ID must exist to use the SNMPv3 security mechanisms (authentication or authentication and encryption) for handling SNMP packets; by default, the engine ID is generated locally. The engine ID can be displayed with the show snmp engineID command, as shown in the following example:
#show snmp engineID SNMP engineID 00000009000000a1ffffffff
Note: If the engineID is changed, all SNMP user accounts must be reconfigured.
The next step is to configure a SNMPv3 group. The following example shows the configuration of a Cisco IOS XR device for SNMPv3 with an SNMP server group AUTHGROUP and enables authentication only for this group by using the auth keyword:
snmp-server group AUTHGROUP v3 auth
The following example shows the configuration of a Cisco IOS device for SNMPv3 with an SNMP server group PRIVGROUP and enables both authentication and encryption for this group by using the priv keyword:
snmp-server group PRIVGROUP v3 priv
The following example shows the configuration of an SNMPv3 user snmpv3user as part of the group snmpusers with a Message-Digest algorithm 5 (MD5) authentication password ofauthpassword and a 3DES encryption password of privpassword:
snmp-server user snmpv3user snmpusers v3 auth md5 authpassword priv 3des privpassword
Protect SNMP Private Community Strings
SNMP private community strings should use complex expressions to prevent guessing or hacking by attackers. For Cisco IOS XR Software Release versions prior to 3.8, the following SNMP configuration must be added to prevent SNMP private community strings from being retrieved:
snmp-server view novacm internet included snmp-server view novacm internet.6.3.16 excluded snmp-server community public view novacm RO
Logging Best Practices
Event logging provides users visibility into the operation of a Cisco IOS XR device and the network on which the device is deployed. Cisco IOS XR Software provides several flexible logging options that can help achieve the network management and visibility goals of an organization.
The following sections provide some basic logging best practices to help administrators utilize logging successfully while minimizing the impact of logging on a Cisco IOS XR device.
See the Implementing Logging Services section of the Cisco IOS XR System Monitoring Configuration Guide for more information about logging configuration.
AAA Logging
AAA logging collects information about user dial-in connections, logins, logouts, HTTP access, privilege-level changes, commands executed, and similar events. AAA log entries are sent to authentication servers using the TACACS+ and RADIUS protocols and are recorded locally by those servers, typically in disk files. If using a TACACS+ or RADIUS server, administrators are advised to enable AAA logging of various types; this is done using AAA configuration commands such as AAA accounting. For details, see the Authentication, Authorization, and Accounting section of this guide.
Access Control List Violation Logging
When using access lists to filter traffic, administrators are advised to log some packets that violate the filtering criteria to find out what type of traffic is being sent to the router. Because access list log messages are rate-limited, the performance impact on Cisco IOS XR devices is minimal.
The following ACL logging example is used to log access violation from class A address 10.0.0.0 to host 202.202.202.20 and includes input interface:
ipv4 access-list violation-log 10 deny ipv4 10.0.0.0 0.255.255.255 host 202.202.202.20 log-input 20 permit ipv4 any any !
Logging Correlation
Logging correlation can be used to isolate the most significant root messages for events affecting system performance.
See the Implementing and Monitoring Alarms and Alarm Log Correlation section of the Cisco IOS XR System Monitoring Configuration Guide for more information about logging.
Send Logs to a Central Location
Administrators are advised to send logging information to a remote syslog server to more effectively correlate and audit network and security events across network devices.
Note: Syslog messages are transmitted unreliably by UDP and are transmitted in plain text. For Any protections that a network affords to management traffic, for example, encryption or out-of-band access, should be extended to include syslog traffic.
The following example shows the configuration of Cisco IOS XR device sending logging information to a remote syslog server with the IP address 10.1.1.1:
logging 10.1.1.1
!
Logging Levels
Each log message generated by a Cisco IOS XR device is assigned one of eight severity levels that range from emergencies to debugging. Unless specifically required, administrators are advised to avoid logging at the debugging level. Logging at the debugging level produces an elevated CPU load that can lead to device and network instability.
The global configuration command logging trap level is used to specify the logging messages that are sent to remote syslog servers. The level specified indicates the lowest severity message that is sent. For buffered logging, the logging buffered level command is used.
The following configuration example shows how to send log messages (from informational through emergencies) to the local log buffer.
logging buffered informational
Disable Console or Monitor Sessions
Cisco IOS XR Software allows administrators to send log messages to monitor sessions. Monitor sessions are interactive management sessions in which the EXEC command, terminal monitor, is issued to the console. The use of the terminal monitor command is not recommended because it can increase the CPU load of a Cisco IOS XR device. Instead, administrators are advised to send logging information to the local log buffer that can be viewed by issuing the show logging command.
Administrators can use the global configuration commands logging console disable and logging monitor disable to disable logging to the console and monitor sessions, as shown in the following configuration:
logging console disable logging monitor disable
Buffered Logging
Cisco IOS XR Software supports the use of a local log buffer so that an administrator can view locally generated log messages. The use of buffered logging is recommended over the use of logging to either the console or monitor sessions.
There are two configuration options that are relevant when configuring buffered logging: the logging buffer size and the message severities that are stored in the buffer. The size of the logging buffer is configured with the global configuration command logging buffered size. The lowest severity included in the buffer is configured using the logging buffered severity command. An administrator can view the contents of the logging buffer through the show logging EXEC command.
The following example configures a logging buffer size of 16384 bytes, as well as a logging severity of informational, indicating that messages at levels emergencies through informational will be stored:
logging buffered 16384 logging buffered informational
Caution: Administrators are advised to use caution when increasing the logging buffer size. The buffer size should not be set to large number in comparison to the total memory available on the device.
Configure Logging Source Interface
To provide an increased level of consistency and reliability when collecting and reviewing log messages, administrators are advised to statically configure a source interface from which all syslog messages will be sent. Accomplished with the logging source interface command, statically configuring the logging source interface ensures that the same IP address appears in all logging messages that are sent from an individual Cisco IOS XR device. For added stability, it is advisable to use a loopback interface as the logging source.
The following configuration example illustrates the use of the logging source-interface interface global configuration command to specify that the IP address of the loopback 0 interface be used for all outgoing syslog messages:
logging source-interface loopback 0
Configure Logging Timestamps
The configuration of logging timestamps helps correlate events across network devices. It is important to implement a correct and consistent logging timestamp configuration so that logging data can be correlated. Logging timestamps should be configured to include the date and time to millisecond precision and to include the time zone in use on the device. Syslog servers and routers also need to be synchronized using the same Network Time Protocol (NTP) clock source.
The following example shows the configuration of logging timestamps with millisecond precision within the Coordinated Universal Time (UTC) zone:
service timestamps log datetime msec show-timezone
Alternatively, it is possible to configure a specific local time zone (instead of using UTC) and configure that information to be present in generated log messages. The following example shows the configuration of a device to use the Pacific Standard Time (PST) zone in the log message timestamp:
clock timezone PST -8 service timestamps log datetime msec localtime show-timezone
Configuration Management
Cisco IOS XR Software includes several configuration management features that can be enabled. These features include functionality to back up software and configurations, the ability to rollback the configuration to a previous version, and the ability to create a detailed configuration change log.
Create Software and Configurations Backups
Currently, there are two methods to back up software and configurations for Cisco IOS XR devices: the use of disk backup, also known as "Golden Disk," and the use of the disk mirroring function.
See the Configuring Disk Backups and Disk Mirroring in Cisco IOS XR Software section of the Cisco IOS XR System Management Configuration Guide for more information about software and configuration backup.
Disk Backup
A system backup disk is created when the system files are backed up to a local storage device for the first time. This process formats the selected device and copies the software packages and system configurations to the local storage device, making it possible to boot a system or perform recovery configuration if the primary disk becomes corrupted. The following example shows how to use the disk backup utility to make a copy of the files from disk0: to disk1:
system backup disk0: disk1:
Disk Mirroring
Disk mirroring replicates the critical data on the primary boot device onto another storage device on the same RP, henceforth referred to as the secondary device. If the primary boot device fails, applications continue to be serviced transparently by the secondary device, thereby avoiding a switchover to the standby RP. The failed primary storage device can be replaced or repaired without disruption of service.
Disk mirroring should mirror only critical data on the primary boot device to a secondary storage device. Disk mirroring should not include non-critical data such as logging data. To separate critical data from non-critical data, the disk devices need to be partitioned according to the following list:
- Disk0: is partitioned to disk0: and disk0a:
- Disk1: is partitioned to disk1: and disk1a:
Disk0: and disk1: are used for critical data, while disk0a: and disk1a: are used for logging data and other non-critical data. Before disk mirroring configuration on the RP, the secondary storage device must be partitioned. The following example shows a disk mirroring configuration:
mirror location 0/rp0/cpu0 disk0:disk1:
Exclusive Configuration Change Access
The Cisco IOS XR Exclusive Configuration Change Access feature ensures that only one administrator can make configuration changes to a Cisco IOS XR device at a given time. This feature helps eliminate simultaneous and potentially conflicting changes being made to the device. The following example shows the exclusive configuration:
configure exclusive
Control Plane
The control plane contains the logical group of all routing, signaling, link-state, and other control protocols that are used to create and maintain the state of the network. The control plane includes the following protocols:
- Open Shortest Path First (OSPF)
- Intermediate System – Intermediate System (IS-IS)
- The Border Gateway Protocol (BGP)
- Internet Control Message Protocol (ICMP)
- Label Distribution Protocol (LDP)
- Resource Reservation Protocol (RSVP)
- Protocol Independent Multicast (PIM)
The control plane also includes functions as the Address Resolution Protocol (ARP), the Bidirectional Forwarding Detection (BFD), and the Layer 2 keepalives that maintain interface state.
It is important that management plane and data plane traffic do not adversely affect the control plane. Without proper protection, it is possible for a data plane event, such as a DoS attack, to impact the control plane and cause disruption to the router. In a worst case scenario, a data plane event could cause disruptions to the wider network. The following sections discuss Cisco IOS XR Software security features and configurations that can help protect and ensure the resilience of the control plane.
General Control Plane Hardening
Protection of the control plane of a network device is critical because the control plane ensures that the management and data planes are maintained and operational.
In many cases, disabling the reception and transmission of certain types of messages on an interface can minimize the amount of CPU load that is required to process unnecessary packets.
IP ICMP Redirects
An ICMP redirect message can be generated by a router when the following conditions are met:
- The interface on which the packet enters the router is the same interface on which the packet is routed out.
- The subnet or network of the source IP address is on the same subnet or network of the next-hop IP address of the routed packet.
- The datagram is not source-routed (not a common occurrence).
If these conditions are met, the router forwards the packet to the next hop (by way of the same subnet using the same interface where the packet arrived) and sends an ICMP (Type 5) redirect message back to the sender of the original packet. This behavior allows the sender to bypass the router and forward subsequent packets directly to the destination or to a router closer to the destination.
There are four types of ICMP redirect messages with the ICMP code 0-3. A malicious user can exploit the ability of the router to send ICMP redirects by continually sending packets to the router, forcing the router to respond with ICMP redirect messages that can have an adverse impact on the CPU and the performance of the router.
In Cisco IOS XR Software, ICMP redirects are disabled by default. To enable ICMP redirects, the ipv4redirects command should be used in the interface configuration mode. The use of this functionality should be controlled.
IP Direct-Broadcast
An IP directed-broadcast is an IP packet with the destination address of a particular IP subnet but that originates from a node that is not part of the destination subnet. Dropping IPv4 directed-broadcasts makes routers less susceptible to DoS attacks, especially the "smurf" attack. In such attacks, the attacker sends IP ICMP echo request from a source with a spoofed IP address to a directed-broadcast address. The request causes all the hosts on the targeted subnet to send replies to the spoofed IP source, causing a flood of ICMP response messages. This action could cause the exploited router to become unusable or consume a significant amount of bandwidth.
In a Cisco IOS XR device, directed-broadcast packets are dropped by default. To enable forwarding of IPv4 directed-broadcasts on an interface, use the ipv4 directed-broadcastcommand in interface configuration mode. If an interface requires IP directed-broadcast support, administrators are advised to control the use of the feature.
Disable IPv4 Source Routing
IPv4 source routing takes advantage of either the Loose Source Route with Record Route Layer 3 header options or the Strict Source Route with the Record Route header option to enable the source of the IP datagram to specify the network path that a packet will take. Attackers could use this functionality in an attempt to bypass security controls in the network.
The Cisco IOS XR device discards (drops and does not process) by default all IPv4 datagrams that contain either the Loose or Strict Source Route header option. No configuration is required unless the default behavior needs to be changed.
Note: This default configuration is different from Cisco IOS Software in which IPv4 source route packets are processed by default.
The following configuration example illustrates the use of the global command to disable the previously enabled processing of source route packets:
! no ipv4 source−route !
ICMP Destination Unreachable
The ICMP Destination Unreachable error message is generated by the router to inform the source that the destination is unreachable for some reason. Generating these ICMP (Type 3) messages can increase CPU utilization on the device.
There are 16 types of ICMP destination unreachable messages with the ICMP code 0-15. Certain management functions take advantage of these error messages. For example, trace route relies on receiving the ICMP Port Unreachable (Type 3, code 3) message to determine when it has reached the last hop. Path MTU Discovery relies on the Datagram Too Big, Fragmentation Required Unreachable (Type 3, code 4) message to discover the appropriate Maximum Transmission Unit (MTU) size.
However, a malicious user could exploit the ability of the router to send ICMP unreachable by continually sending packets to the router. An exploit may force the router to respond with ICMP unreachable, resulting in an adverse impact on the CPU and the performance of the router. ICMP unreachable message generation is enabled by default and can be disabled using the interface configuration command
! ipv4 unreachable disable !
When IP unreachable generation is enabled, the generation of these messages is rate-limited by default to one message per 500 milliseconds (ms). However, this default value can be changed by using the global configuration command as shown in the following example:
! icmp ipv4 rate-limit unreachable [DF] < #of messages per ms
!
Cisco IOS XR Software maintains two timers: one for general destination unreachable messages and one for data fragmentation (DF) destination unreachable messages. Both timers share the same time limits and defaults. If the DF keyword is not configured, the icmp ipv4 rate-limit unreachable command sets the time values for DF destination unreachable messages. The DF option limits the rate at which ICMP destination unreachable messages are sent for Type 3, code 4 "packet-too-big, fragmentation is needed and data fragmentation bit is set". If the DF option is configured, its time values remain independent from values of general destination unreachable messages. Without the DF option configuration, all messages would share the same time limits and defaults, as shown in the following configuration:
! icmp ipv4 rate-limit unreachable 1000 !
IPv4 Options Packets
IPv4 packets that contain header options present two security concerns. First, such traffic must be process-switched by Cisco IOS XR devices, resulting in higher CPU usage (that is, the software path using the main CPU of the card is used instead of the hardware path forwarding). Second, packets with IPv4 source route header options can alter the path that traffic takes through the network, as previously described in the IPv4 Source Routing section. This behavior could allow such packets to subvert security controls.
In Cisco IOS XR Software, all IPv4 packets with any header option other than the "source-route" header options that are dropped by default (as detailed in the Disable IPv4 Source Routing section of this guide) are punted, police rate-limited, and processed by ucode and static policers depending on the destination address of the packets. In this way, Cisco IOS XR software prevents packets with IPv4 header options from overwhelming or elevating the CPU load by default.
Note: The rate-limit value is not configurable.
The show ipv4 traffic EXEC command can be used to determine the number of IPv4 options packets that are received and processed.
See the Control Plane section of this guide for more information about LPTS and static policers.
IPv4 Fragments Rate Limiting
Fragmented IPv4 packets can pose a challenge to network devices. Fragmentation is often used in attempts to evade detection by ACLs, intrusion protection systems, and firewalls. Additionally, when the router is the destination of the fragmented IPv4 packet, these packets can be punted to the LC CPU and cause high CPU usage. For these reasons, IPv4 fragments are often used in DoS attacks.
In Cisco IOS XR Software versions 3.6 and later, configurable LPTS hardware policers are used to set the fragment packets default rate to limit the number of fragmented packets that can be punted to the LC CPU, which helps to reduce LC CPU usage.
The following configuration example illustrates the use of LPTS configurable policers:
! lpts pifib hardware police locationflow fragment rate 10 !
Note: Administrators are advised to set IP fragment rates as low as possible. The current Cisco IOS XR device default fragment rate is 1000 p/s. This rate can cause significant LC CPU utilization when receiving a high volume of fragment packets. For Cisco IOS XR Software Release versions prior to 3.6, administrators should contact the Cisco Technical Assistance Center if default LPTS hardware police values require adjustment.
IPv4 Fragments Filtering
The filtering of fragmented IP packets can pose a challenge to security devices, because the Layer 4 information that is used to filter TCP and UDP packets is present only in the initial fragment.
Cisco IOS XR Software checks non-initial fragments of the packets by evaluating them against the ACL and ignoring any Layer 4 filtering information. This behavior causes non-initial fragments to be evaluated solely on the Layer 3 portion of any configured access control entry (ACE).
In the following example configuration, if a TCP packet destined for 192.168.1.1 on port 22 is fragmented in transit, the initial fragment is dropped as expected by the second ACE based on the Layer 4 information within the packet. However, all remaining (non-initial) fragments are allowed by the first ACE based on the Layer 3 information in the packet and ACE. ipv4 access-list ACL-FRAGEMENT EXAMPLE
! ipv4 access-list ACL-FRAGEMENT EXAMPLE ! 10 permit tcp any host 192.168.1.1 eq 80 20 deny tcp any host 192.168.1.1 eq 22 !
Due to the non-intuitive nature of fragment handling, IP fragments are often inadvertently permitted by ACLs. Fragmentation is also often used in attempts to evade detection by intrusion detection systems. It is for these reasons that IP fragments are often used in attacks, and why they must be explicitly filtered at the top of any configured ACLs. The following example ACL includes comprehensive filtering of IP fragments. The functionality in the following example must be used in conjunction with the functionality in the preceding example.
! ipv4 access-list ACL-INFRASTRUCTRE-IN ! !---Deny IP fragments using protocol-specific ACEs to aid in !---classification of attack traffic ! 10 deny tcp any any fragments 20 deny upd any any fragments 30 deny icmp any any fragments 40 deny ipv4 any any fragements ! !---Deny all other IP traffic to any network device ! 50 deny ipv4 any! !—Permit transit traffic ! 60 permit ipv4 any any !
See the Access Control Lists and IP Fragments white paper for more details about advanced configuration.
See the Extended Access Control Lists with Fragment Control section of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information about ACL handling of fragmented IP packets.
IPv6 ICMP Rate Limiting
IPv6 ICMP error messages are generated by the router for many reasons. The generation of IPv6 ICMP error messages is enabled by default, but the rate at which these messages are generated can be rate-limited. This action is accomplished by using the global configuration command.
! ipv6 icmp error-interval milliseconds [bucketsize] !
The IPv6 ICMP rate limiting feature implements a token bucket algorithm and limits the rate at which IPv6 ICMP error messages are sent out on the network. The initial Cisco IOS XR Software implementation of IPv6 ICMP rate limiting defined a fixed interval between error messages. However, some applications, such as trace route, often require replies to a group of requests that are sent in rapid succession. The fixed interval between error messages is not flexible enough to work with applications such as trace route and can cause the application to fail. Implementing a token bucket scheme allows a number of tokens—representing the ability to send one error message each—to be stored in a virtual bucket. The maximum number of tokens allowed in the bucket can be specified, and for every error message sent, one token is removed from the bucket. If a series of error messages is generated, error messages can be sent until the bucket is empty. When the bucket becomes depleted of tokens, IPv6 ICMP error messages will not be sent until a new token is placed in the bucket. The token bucket algorithm does not increase the average rate limiting time interval and is more flexible than the fixed time interval scheme. The following sample configuration shows how to change the interval time:
! ipv6 icmp error-interval 50 20 !
IPv6 Packet with Header Extension
By default, Cisco IOS XR Software only punts to the CPU for handling IPv6 packets that contain the Hop-by-Hop header extension (NH=0). Currently, Cisco IOS XR software does not process IPv6 Routing Header Extensions (NH=43) and others. The default and only available mode of operation is to drop IPv6 packets with routing headers.
Note: Cisco IOS XR Software does not distinguish between Type 0 (source-route) and Type 2 (Mobile IP) Routing Header Extensions.
Punted IPv6 header extension packets are rate-limited by ucode in hardware. In this way, Cisco IOS XR Software prevents packets with IPv6 header extension from overwhelming or elevating the CPU load.
Note: The rate-limit value is not configurable.
TCP Service and Accept Rate Limiting
Most TCP services are disabled by default in Cisco IOS XR Software. After enabling a particular TCP service, a restriction of the incoming traffic rate should be enabled using MPP, peer control, or ACL.
Another way of restricting TCP traffic is to limit the number of accepted incoming TCP connections per second. Doing so can help mitigate DoS attacks from trusted neighbors and reduce CPU usage. The current default rate is 500 connections per second; this number should be carefully tuned on a case-by-case basis.
The following example shows how to reduce the number of accepted TCP connections to 200 connections per second:
! tcp accept-rate 200 !
Proxy Address Resolution Protocol
Proxy address resolution protocol (ARP) is the technique in which one device, usually a router, answers ARP requests that are intended for another device. By acting on behalf of the other device, the router accepts responsibility for routing packets to the real destination. Proxy ARP can help networked devices on one subnet reach devices on remote subnets without configuring routing or a default gateway. Proxy ARP is defined in RFC 1027.
There are several disadvantages to using proxy ARP. The use of proxy ARP can result in increased ARP traffic on the network, segment and resource exhaustion, and man-in-the-middle attacks. Proxy ARP can be used by a resource exhaustion attack vector because each proxy ARP request consumes a small amount of memory. An attacker could exhaust all available memory by sending a large number of ARP requests. Man-in-the-middle attacks enable a host on the network to spoof the MAC address of the router, resulting in unsuspecting hosts sending traffic to the attacker.
The proxy ARP function is disabled by default in Cisco IOS XR software. To enable it, a proxy-arp command must be used on the interface configuration mode. If the proxy ARP function is absolutely required, administrators should control its use.
Network Time Protocol
Maintaining accurate time is essential for many aspects of network operations and security. The Network Time Protocol (NTP) is a UDP-based protocol that provides the ability to accurately maintain consistent time across a large number of network devices. NTP is especially useful to ensure that timestamps on log messages are consistent throughout the entire network. To prevent security attacks, NTP protocol should always use trusted time sources and proper authentication.
Two mechanisms for securing NTP are available: an access list-based restriction scheme and an encrypted authentication mechanism.
- Access list-based restriction: The access list-based restriction scheme allows administrators to grant or deny certain access privileges to an entire network, a subnet within a network, or a host within a subnet. The following example shows access list-based restriction using access group:
(config)# ntp (config-ntp)# peer 10.1.1.1 (config-ntp)# peer 10.2.2.2 (config-ntp)# access-group peer peer-acl (config-ntp)# access-group serve serve-acl (config-ntp)# access-group serve-only serve-only-acl (config-ntp)# access-group query-only query-only-acl
- NTP Authentication: NTP authentication configuration provides the assurance that all NTP messages are exchanged between trusted NTP peers and that their content has not been modified. The following example shows an NTP authentication configuration:
NTP Client
(config)# ntp (config-ntp)# authenticate (config-ntp)# authentication-key 5 md5 encrypted ciscotime (config-ntp)# trusted-key 5 (config-ntp)# server 172.16.1.5 key 5NTP Server
(config)# ntp (config-ntp)# authenticate (config-ntp)# authentication-key 5 md5 encrypted ciscotime (config-ntp)# trusted-key 5
Note: The encryption and decryption processes used in NTP authentication can be very CPU-intensive and can seriously degrade the accuracy of the time that is propagated within a network. On networks that permit a more comprehensive model of access control, administrators should consider using the access-list-based form of control instead.
NTP services are disabled on all interfaces in Cisco IOS XR Software by default. Administrators should enable it only on the specific interface when necessary. When NTP is enabled globally, administrators can selectively prevent NTP packets from being received through a specific interface by turning off NTP on a given interface, as shown in the following example:
(config)# ntp (config-ntp)# no interface pos 0/0/0/1
or
(config)# ntp (config-ntp)# interface POS 0/0/0/1 disable
Limit CPU Impact of Control Plane Traffic
Protection of the control plane is critical. Because application performance and end-user experience can suffer without the presence of data and management traffic, the survivability of the control plane ensures that the other two planes are maintained and operational.
Control Plane Traffic
Under normal network operating conditions, the vast majority of packets handled by network devices are data plane packets, and routers are optimized to forward these packets efficiently in the hardware and fast paths. However there are certain types of packets that must be punted to be handled directly by the LC or RP CPUs. The types of packets in this group can be divided into the following categories.
Note: Packets that are listed in italics include traffic that is punted to RP CPU. The packets listed in straight text are processed directly by the LC CPU.
- Transit traffic: IP option, IP option router alert, ARP resolution, and TTL expired packets
- Unicast receive (to us) traffic: Various ICMP, management traffic like SSH, SNMP and XML, NetFlow, and Cisco DP packets along with routing packets such as OSPF, BGP, and ISIS
- Multicast and Broadcast: Multicast control traffic like PIM and HSRP, broadcast packets, and the first packet of MCAST stream
- Special Traffic: IP and MPLS fragments and L2 packets
The preceding traffic types can be a potential source of DoS attacks and the system must have the ability to police such flows.
Local Packet Transport Services
As routers handle greater traffic loads and networks become larger and more complex, it is increasingly difficult for network engineers to manually configure all of the controls necessary to protect the router and core network infrastructure. The Cisco IOS XR Software directs router self-protection beyond manual configuration and the mere detection and reporting on security violations. Cisco IOS XR Software provides intelligence that automatically provisions many aspects of router self-protection functions, including differentiating between unidentified and trusted control plane sessions. In a Cisco IOS XR device, this functionality is provided by hardware rate-limiters that are associated with LPTS. The primary functions of LPTS include the following attributes:
- Ability to control packet flows for all internal applications that receive packets from outside the router.
- Exceptions packet rate-limiters that control all other packets that require punting for CPU support.
- Preconfigured LPTS and exceptions packet rate-limiters for use with default values; these rate-limiters are capable of functioning without the need for any user-configuration.
- Static programming of LPTS policers during boot; these policers are automatically applied based on the flow type of the incoming traffic. This automation frees time spent by network administrators on manual configuration for use on other mission-critical tasks.
- User capability to configure customized rate-limits in the event that the default policer settings are inadequate for the specific deployment model. The user can configure the rate per policer per node (locally) or globally from CLI, thus overwriting default static policer values. This attribute applies to Cisco IOS XR Software Release versions 3.6 and later.
The command that is used to configure LPTS police for a particular flow type is as follows:
[no] lpts pifib hardware police [location]
If the "location" option is not specified, it is equivalent to the "location all" option in the configuration command, causing the command to be applied on all nodes.
For example, BGP-known traffic is controlled in the following configuration:
! lpts pifib hardware police flow bgp-known rate 20000 !
Default LPTS policer values and flow-type values can be seen using the following show command:
!
show lpts pifib hardware police [location {all | }]
! In addition to LPTS, which controls the flow of all receive packets, a separate group of hardware-based rate-limiters automatically control "exceptions" along with transit packets that require punting for CPU support. Examples of packets of this type include various Layer 2 packets, Cisco DP, IPv4 TTL errors, IPv4 options, ARP and RARP, and many others. Because these packet types do not rep resent traffic flows that the router wants to keep track of, there is no need to provide per flow control, which is the case with LPTS and "receive" packets. Therefore, these "exceptions" rate-limiters are simply programmed statically and are not user-configurable.
See the Implementing LPTS section of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information.
CPU and Routing Protocol Software Queues
When properly configured, LPTS policers should mitigate most cases where overflowing the line card (LC) or the routing protocol (RP) CPU may occur. Cisco IOS XR Software implements additional mechanisms in case the amount of punted packets exceeds the processing capability of the CPU. Before the packets reach a CPU, they are placed in one of a number of software queues that perform priority queuing should congestion occur.
Four queues that classify packets are available for LC CPU as shown in the following:
- Critical: Layer 2 keep alive (PPP, HDLC)
- High: ARP
- Medium: IPv4 lookup
- Low: TTL errors, Options, logging, ICMP
Three queues that classify packets are available for the RP as seen in the following list:
- High: OSPF, ISIS
- Medium: BGP, PIM, LDP, SSH
- Low: Fragments
General Routing Protocol Securing Techniques
Multiple Cisco IOS XR features have been developed to protect IP protocols against various attacks. The following section details protocol-independent features. The sections that follow will focus on the specific examples of the protocol implementation security features.
Message-Digest Algorithm 5 Peer Authentication
Message-Digest algorithm 5 (MD5) is a widely used cryptographic hash function with a 128-bit hash value. Most routing protocols can use MD5 to generate a hash-based MAC. MAC provides an integrity check to protect control packets that are exchanged between routing protocol session peers. Each protocol packet carries an MD5 digest message. This digest acts like a signature for that segment, incorporating information that is known only to the connection end points. Using MD5 authentication significantly mitigates certain types of attacks against routing protocols. MD5 should be configured on each session. Protocols like BGP, OSPF, ISIS, and LDP have the option to configure MD5 to secure their sessions.
Keychain Management
Routing protocols and network management applications often use authentication for enhanced security while communicating with their peers. Most authentication methods use shared secrets, sometimes referred to as keys, on all the entities in their mechanisms for establishing trust between each peer.
As with all key-based security methods, it is useful to specify a lifetime for each key so that these keys are changed on a regular basis when they expire. To maintain stability during such key changes, each party must be able to store and use more than one key for an application at the same time. The sequence of keys intended for authenticating the same peer or peer group is collectively managed in a container called a key chain. Each key in the key chain has an associated lifetime. Multiple cryptographic algorithms options can also be specified.
A keychain is defined as a set of keys that each include an associated lifetime, authentication-algorithm, and key-id. Hitless key rollover provides a switchover from one (expired) key to another (active) key, without any session-drops. Border Gateway Protocol (BGP), Open Shortest Path First (OSPF), and Intermediate System-to-Intermediate System (IS-IS) use the keychain to implement a hitless key rollover for authentication.
The following is an example of the keychain configuration:
! key chain my-key-chain accept-tolerance key 8 key-string mykey_test cryptographic-algorithm MD5 send-lifetime 1:00:00 june 29 2009 infinite accept-lifetime 1:00:00 june 29 2009 infinite !
See the Implementing Keychain Management section of the Cisco IOS XR System Security Configuration Guide for more information about keychain management.
Routing Policy Language
In Cisco IOS XR Software, a new routing policy language (RPL) has been designed to provide a single, straightforward language in which all routing policy needs can be expressed. RPL was designed to support large-scale routing configurations and thus can be used by BGP and by IGP protocols such as OSPF and ISIS. RPL replaces the route map statements that are used in Cisco IOS Software.
The policy language introduces the notion of route policies and sets. Sets are containers of similar data that can be used in route attribute matching and setting operations. Four set types exist: prefix-sets, community-sets, as-path-sets, and ext community-sets. These sets hold groupings of IPv4 or IPv6 prefixes, community values, autonomous system (AS) path regular expressions, and extended community values, respectively.
Route policies use the sets to instruct the router to inspect and filter routes and potentially modify route attributes when they are accepted from a peer, advertised to a peer, or redistributed from one routing protocol to another. Routing protocols make decisions to advertise, aggregate, discard, distribute, export, hold, import, redistribute, and otherwise modify routes based on configured routing policy.
Route policies can be attached to different attach points when an association is formed between a specific protocol entity, in this case, a BGP neighbor, and a specific named policy.
RPL is a very powerful tool and this document is not intended to describe the whole language in detail. For more information about RPL, see the Implementing Routing Policy section of the Cisco IOS XR Routing Configuration Guide.
Secure Border Gateway Protocol
The Border Gateway Protocol (BGP) is the routing foundation of the Internet. As such, any organization with more than modest connectivity requirements often finds itself utilizing BGP. BGP is often targeted by attackers because of its ubiquity and the "set and forget" nature of BGP configurations in smaller organizations. However, there are many BGP security features that can be leveraged to increase the security of a BGP protocol.
This section provides an overview of the most important BGP security features. Where appropriate, configuration recommendations are made.
See the Implementing BGP section of the Cisco IOS XR Routing Configuration Guide for more information.
BGP Time-to-Live-Based Security Protection
The BGP time-to-live (TTL) -based security check is designed to protect BGP processes from CPU utilization-based attacks. The BGP protocol defines two types of sessions: "internal" BGP sessions (iBGP) that are established between peers within the same Autonomous System (AS), and external BGP (eBGP) sessions that are established between peers in two different ASs. Of particular interest are the eBGP sessions, which are the type of BGP sessions that would be established between an Enterprise and its upstream service provider.
By default and per RFC 3682, when eBGP is configured, the IP header TTL for all neighbor session packets is set to "1". This configuration was originally assumed to be useful because it prevented the establishment of an eBGP session beyond a single hop; however, the configuration does not take into account an attacker who could be located up to 255 hops away and could have the ability to send spoofed packets to BGP-speaking routers. For example, an attacker could send large amounts of TCP SYN packets at the BGP peer to overwhelm the BGP process. The BGP TTL security check leverages ISP eBGP peering sessions between routers that are adjacent to each other (either between directly connected interfaces or possibly between loopbacks). Because TTL spoofing is considered nearly impossible, a mechanism that is based on an expected TTL value was developed to provide a simple yet robust defense from infrastructure attacks that are based on forged BGP packets.
Note: This concept was originally defined in the The BGP TTL Security Hack (BTSH) document and later extended beyond BGP in RFC 3682, The Generalized TTL Security Mechanism (GTSM).
When the BGP TTL security check is enabled, the initial value for eBGP is set to "255" (instead of "1"), and a minimum TTL value is enforced on all BGP packets that are associated with that eBGP session. Because the IP header TTL value is decremented by each hop (router interface) along its path to its final destination, using BTSH restricts possible attack origins to the directly connected routers.
Currently, Cisco IOS XR Software does not support GTSM for eBGP multi-hop scenarios; only directly connected or loopback-based eBGP peering sessions can use the GTSM feature. This feature is checked in hardware during packets ingress in LCs. GTSM for Cisco IOS XR BGP can be enabled using the following commands:
! router bgpneighbor ttl-security !
BGP Peer Authentication with MD5
BGP neighbor sessions are established between two peers and then routes are exchanged between each other. By enabling the MD5-based neighbor authentication mechanism, administrators can ensure that only authorized peers establish this BGP neighbor relationship, and that the routing information exchanged between these two devices has not been altered en-route between the two systems.
The BGP neighbor authentication process is a symmetric technique. Therefore, when this process is turned on, it must be simultaneously enabled on both sides of the peering session. Neighbor authentication using MD5 creates an MD5 digest hash for each packet that is sent as part of a BGP session. Specifically, portions of the IP and TCP headers, TCP payload that includes the BGP route advertisements, and the shared secret key, are used to generate the digestMD5 hash by the sending router. The created digest MD5 hash is then stored in TCP option Kind 19, which was created specifically for this purpose in the RFC 2385. The receiving BGP speaker neighbor uses the same algorithm and shared secret key to regenerate and compute its own version of the message digestMD5 hash. The receiving BGP speaker neighbor compares its own version with the one it received; if the received and computed digestsMD5 hash values are not identical, the packet is discarded. Otherwise, the packet is accepted and processed by BGP.
Peer authentication with MD5 is configured by using the password option to the neighbor BGP router configuration command. The following example illustrates the use of this command:
! router bgpneighbor remote-as password {clear | encrypted} password !
See RFC 2385, Protection of BGP Sessions via the TCP MD5 Signature Option, for more information about BGP MD5 peering authentication.
BGP Peer Authentication with Keychain
BGP uses TCP authentication, which enables the authentication option and sends the MAC based on the cryptographic algorithm configured for the keychain.
The following example shows how to apply the keychain between BGP neighbors:
! router bgpneighbor remote-as keychain !
BGP Maximum Prefixes
BGP prefixes are stored by a router in memory. The more prefixes a router must hold, the greater the memory consumed by BGP. Both malicious and inadvertent (misconfiguration) processes can cause excessive prefixes to be seen by a BGP peer. To prevent memory exhaustion, it is important to configure the maximum number of prefixes accepted on a per-peer basis. Administrators are advised to configure limits for each BGP peer.
When configuring BGP prefixes using the neighbor maximum-prefix BGP router configuration command, one argument is required: the maximum number of prefixes that are accepted before a peer is shutdown. Optionally, a number from 1–100 can also be entered. This number represents the percentage of the maximum prefixes value at which point a log message can be sent. The following is an example of a BGP configuration:
! router bgpneighbor remote-as address-family unicast maximum-prefix [threshold] [warning-only] !
The BGP maximum-prefix feature allows users to control the number of prefixes that can be installed from a neighbor. When configured, this feature will bring down a neighbor relationship when the number of received prefixes from that peer exceeds the configured maximum-prefix limit. Typically, the maximum-prefix threshold would be set with some margin of error over a target value to accommodate some degree of unexpected changes in topology. The other warning-only option does not cause the neighbor relationship to be brought down, but simply issues a message when the threshold is exceeded. This feature is commonly used for external BGP peers but can also be applied to internal BGP peers.
Filter BGP Prefixes with RPL policies
In Cisco IOS XR devices, BGP allows users to filter prefixes that are based on network prefixes, as-paths, and community or ext-community values.
Prefix sets allow a network administrator to permit or deny specific prefixes that are sent or received by way of BGP. Prefix sets should be used when possible to ensure that network traffic is sent over the intended paths. Prefix lists should be applied to each eBGP peer in both the inbound and outbound directions.
Configured prefix sets limit the prefixes that are sent or received to prefixes specifically permitted by the routing policy of a network. If this is not feasible because of a large number of prefixes received, a prefix set should be configured to specifically block known bad prefixes. These known bad prefixes include unallocated IP address space and networks that are reserved for internal or testing purposes by RFC 3330. Outbound prefix lists should be configured to specifically permit only the prefixes that an organization intends to advertise.
The following is a simple example of the BGP route policy using a prefix set to drop 10.0.2.0/24 and 10.0.3.0/24 incoming prefixes. It uses prefix-set and the policy is attached to the neighbor inbound attach point.
! prefix-set bgp_route_drop 10.0.2.0/24, 10.0.3.0/24 end-set ! route-policy bgp_in if destination in bgp_route_drop then drop endif end-policy ! router bgp < as-number > neighbor < address > address-family ipv4 unicast route-policy bgp_in in !
The following configuration uses as-path-set to permit inbound prefixes, which include AS path 12 or 13, but to drop other prefixes:
! as-path-set my-as-set ios-regex ‘12$’, ios-regex ‘13$’ end-set route-policy policy-a if as-path in my-as-path then pass else drop endif end-policy ! router bgp < as-number > neighbor < address > address-family ipv4 unicast route-policy policy-a in !
For more information about RPL BGP, see the Implementing Routing Policy section of the Cisco IOS XR Routing Configuration Guide.
Secure Interior Gateway Protocols
The ability of a network to properly forward traffic and recover from topology changes or faults is dependent on an accurate view of the topology. Running an IGP can often provide this view. By default, IGPs are dynamic and discover additional routers that communicate with the particular IGP in use. IGPs also discover routes that can be used during a network link failure.
The following subsections provide an overview of the most important IGP security features. Recommendations and examples that cover OSPF, ISIS, Routing Information Protocol Version 2 (RIPv2), and Enhanced Interior Gateway Routing Protocol (EIGRP) are provided when appropriate.
See the Cisco IOS XR Routing Configuration Guide for more information about IGP protocol configuration.
IGP TTL Security
The Generalized TTL Security Mechanism (GTSM) (RFC 3682) is designed to protect the control plane of a router from CPU utilization-based attacks. GTSM recognizes that the vast majority of protocol peerings are established between routers that are adjacent to or, in a worst case scenario, between loopbacks on adjacent routers. Because TTL spoofing is considered nearly impossible, a mechanism that is based on an expected TTL value can provide a simple and reasonably robust defense from infrastructure attacks that are based on forged protocol packets. GTSM is equally applicable to both TTL (IPv4) and Hop Limit (IPv6).
The initial Cisco IOS XR Software implementation for GTSM supported BGP peering protection; see the section of this guide for more information. Cisco IOS XR Software provides extended support for GTSM to cover the OSPF IGP.
OSPF is a link state protocol that requires networking devices to detect topological changes in the network; flood Link State Advertisement (LSA) updates to neighbors; and quickly converge on a new view of the topology. However, while receiving LSAs from neighbors, network attacks can occur, because no checks occur to distinguish whether unicast or multicast packets are originating from a neighbor that is one hop away or multiple hops away over virtual links. For virtual links, OSPF packets travel multiple hops across the network; therefore, the TTL value can be decremented several times. For virtual links, a minimum TTL value must be allowed and accepted for multiple hop packets.
The following is a sample configuration of GTSM for OSPF. GTSM can be configured on a per-interface basis or globally for all OSPF peers. The number of hops is the parameter that changes the default expected TTL value of incoming packets.
! router ospf 1 area 1 interface GigabitEternet0/5/0/0 security ttl hops 2 !
IGP Peer Authentication with MD5
Failure to secure the exchange of routing information could allow an attacker to introduce false routing information into the network. By using password authentication with routing protocols between routers, administrators can help secure the network. Because this authentication is sent as plain text; however, the attacker could easily subvert this security control. By adding MD5 hash capabilities to the authentication process, the routing updates no longer contain plaintext passwords, and the entire contents of the routing update are more resistant to tampering.
MD5 authentication is still susceptible to brute force and dictionary attacks if weak passwords are chosen. Administrators are advised to use passwords with sufficient randomization. MD5 authentication is more secure than password authentication
The following subsections detail the specific implementations of the OSPF and IS-IS protocols.
OSPF Authentication with MD5
All OSPF routing protocol exchanges are authenticated and the method used can vary depending on how authentication is configured. When using cryptographic authentication, the OSPF routing protocol uses the MD5 authentication algorithm to authenticate packets that are transmitted between neighbors in the network. For each OSPF protocol packet, a key is used to generate and verify a message digest that is appended to the end of the OSPF packet. The message digest is a one-way function of the OSPF protocol packet and the secret key. Each key is identified by the combination of the interface used and the key identification.
The following configuration example shows OSPF authentication with MD5 Configuration:
! router ospf 1 area 1 authentication message-digest message-digest-key 100 md5 cisco !
Intermediate System to Intermediate System Authentication with MD5
Authentication can be configured to limit the establishment of adjacencies by using the hello-password command, and limit the exchange of LSPs by using the lsp-password command.
Intermediate System to Intermediate System (IS-IS) supports plaintext authentication, which does not provide security against unauthorized users. The password is exchanged as plain text and is potentially visible to an agent who can view the IS-IS packets. When an HMAC-MD5 password is configured, the password is never sent over the network and, instead, is used to calculate a cryptographic checksum to ensure the integrity of the exchanged data.
To set the domain password, configure the lsp-password command for Level 2; to set the area password, configure the lsp-password command for Level 1.
The following configuration example shows IS-IS authentication with MD5 configuration:
! router isis Core lsp-password hmac-md5 cisco123 interface GigabitEternet0/5/0/0 hello-password hmac-md5 cisco123 !
Open Shortest Path First Authentication with Keychain
The following examples show the OSPF configuration, which establishes a keychain to set the password at the global, area, and interface levels.
Router-level configuration:
! router ospf 1 router-id 10.2.3.4 authentication message-digest keychain my-key-chain !
Area-level configuration:
! router ospf 1 router-id 10.2.3.4 area 0 authentication message-digest keychain my-key-chain !
Configuration at virtual-link level:
! router ospf 1 router-id 10.2.3.4 area 0 virtual-link 1.1.1.1 authentication message-digest keychain my-key-chain !
Interface-level configuration:
! router ospf 1 router-id 10.2.3.4 area 0 interfaceauthentication message-digest keychain my-key-chain !
IS-IS Authentication with Keychain
The following example shows IS-IS configuration, which configures a keychain to set the domain and area passwords at the global and interface levels.
! router isis 1 lsp-password keychain isis-keys level-1 interfacehello-password keychain isis-keys !
Passive Interface
The use of the passive interface capabilities within IGPs has two benefits. First, such capabilities reduce the CPU load on the router by suppressing the generation and processing of IGP protocol messages on the identified interface(s). Second, these capabilities could help mitigate information leaks by not advising IGP protocol information to networks beyond administrative control.
The following example shows the use of the passive command for the OSPF protocol. If the passive command is not configured, the interface generates and receives normal OSPF traffic flow (default behavior).
! router ospf 1 area 0 interface Loopback0 passive !
The following shows the ISIS ISIS configuration example:
! router isis 1 interface Loopback0 passive !
The RIP RIP configuration example follows:
! router rip interface Loopback0 passive-interface !
The EIGRP EIGRP configuration example follows:
! router eigrp 1 interface Loopback0 address-family ipv4 passive-interface !
IGP Route Filtering
Similar to BGP, the IGP protocols also benefit from having the ability to filter ingress and egress routing information and can use the Cisco IOS XR Software RPL language for this purpose.
See the Implementing Routing Policy section of the Cisco IOS XR Routing Configuration Guide for more information.
IGP Routing Process Resource Consumption
Routing Protocol prefixes are stored by a router in memory, and resource consumption increases with the additional prefixes that a router must hold. To prevent resource exhaustion, it is important to set protocol limits for number of accepted and stored prefixes.
See the Implementing Routing Policy section of the Cisco IOS XR Routing Configuration Guide for more information.
OSPF Maximum Redistributed Prefix Limit
To limit the aggregate number of routes that may be redistributed into an OSPF process, administrators are advised to use the maximum redistributed-prefix command in the appropriate mode. The following example shows how to configure a maximum number of routes that can be redistributed for an OSPF routing process:
! router ospf 10 maximum redistributed-prefixes 15000 !
IS-IS Redistributed Prefix Limit
To specify an upper limit on the number of external prefixes (subject to summarization) that the Intermediate System-to-Intermediate System (IS-IS) protocol advertises, use the maximum-redistributed-prefixes command in address family mode. The following example shows how to specify the number of redistributed prefixes at 5000 for Level 2:
! router isis isp address-family ipv4 unicast maximum-redistributed-prefixes 5000 level 2 !
Enhanced Interior Gateway Routing Protocol Redistributed Prefix Limit
To limit the number of prefixes that are redistributed to an Enhanced Interior Gateway Routing Protocol (EIGRP) process, use the redistribute maximum-prefix command in the IPv4 virtual routing and forwarding (VRF) address family configuration mode. The following example shows how to configure the maximum prefix limit for routes that are learned through redistribution under the vrf vpn-1 command. The maximum limit is set to 5000 prefixes and the warning threshold is set to 95 percent. When the number of prefixes that is learned through redistribution reaches 4750 (95 percent of 5000), warning messages are displayed in the console. Because the warning-only keyword is configured, the topology and routing tables are not cleared and route redistribution is not placed in a penalty state.
! router eigrp 100 vrf vpn-1 address-family ipv4 redistribute maximum-prefix 5000 95 warning-only !
Secure Label Distribution Protocol
Label Distribution Protocol (LDP) performs label distribution in MPLS environments. As LDP assigns and distributes labels, the stability of this protocol is a critical factor for reliable MPLS networks. LDP uses the hello discovery mechanism to discover its neighbors and create sessions between them. To secure data transmitted over these sessions, password authentication (using the TCP MD5 option) should be configured for a given neighbor using password command as shown below:
! mpls ldp … neighborpassword encrypted 110A1016141E !
Secure Resource Reservation Protocol
The Resource Reservation Protocol (RSVP), as described in RFC 2205, is a Transport layer protocol that is designed to reserve resources across the network for integrated services. RSVP provides a receiver-initiated setup of resource reservations for multicast or unicast data flows. An extension of RSVP for traffic engineering is used for the MPLS TE tunnels setup.
Secure RSVP with ACL
Cisco IOS XR implementation of RSVP enables ACLs to forward, drop, or perform processing on RSVP Router-Alert (RA) packets. For each incoming RSVP RA packet, RSVP inspects the IP header and attempts to match the source and destination IP addresses with a prefix that is configured in an extended ACL. In the following example, rsvp_acl_example ACL is configured to permit RA packets with the source address 1.1.1.1 and drop packets with the source address 2.2.2.2. All other RA packets are processed as normal RSVP packets.
! ipv4 access-list rsvp_acl_example 10 permit ipv4 hostany 20 deny ipv4 any host ! rsvp signalling prefix-filtering access-list rsvp_acl_example !
The default behavior of Cisco IOS XR Software will perform normal RSVP processing on RA packets when the ACL match returns an implicit (default) deny. To change the default behavior, the following command must be issued:
! signalling prefix-filtering default-deny-action drop !
RSVP Authentication
RSVP authentication supports only keyed-hash message authentication code (HMAC) type algorithms. The RSVP authentication feature permits neighbors in an RSVP network to use a secure hash to sign all RSVP signaling messages digitally, thus allowing the receiver of an RSVP message to verify the sender of the message without relying solely on the IP address of the sender. The signature is accomplished on a per-RSVP-hop basis with an RSVP integrity object in the RSVP message, as defined in RFC 2747.
RSVP authentication can be configured in global, interface, or neighbor configuration modes. In global mode, a single common key set is expected to be used to authenticate all RSVP messages, while in the other modes, different keys can be used. A keychain has to be configured first to enable authentication on the RSVP. See the Keychain Management section of this guide for more details about keychain creation. The following configuration shows how to enable RSVP authentications on different modes:
! rsvp authentication key-source key-chain default_keys ! interfaceauthentication key-source key-chain default_keys ! neighbor authentication key-source key-chain nbr_keys !
Secure First Hop Redundancy Protocols
First Hop Redundancy Protocols (FHRPs) provide resiliency and redundancy for devices that are acting as default gateways. This situation and these protocols are commonplace in environments where a pair of Layer 3 devices provides default gateway functionality for a network segment or set of VLANs that contain servers or workstations.
The Hot Standby Router Protocol (HSRP) and the Virtual Router Redundancy Protocol (VRRP) are examples of FHRPs. By default, these protocols communicate using unauthenticated packets. This kind of communication could allow an attacker to pose as an FHRP-speaking device to assume the default gateway role on the network. This takeover would allow an attacker to perform a man-in-the-middle attack and intercept all user traffic that exits the network.
VRRP Text Authentication
Text authentication can ensure that VRRP messages received from adjacent routers that comprise a virtual router are authenticated by configuring a simple text password, as shown in the following configuration:
! router vrrp interfacevrrp 1 text-authentication !
HSRP Text Authentication
Similar to VRRP, HSRP authentication can be configured using the following authentication command:
! router hsrp interfacehsrp 1 authentication !
See the Implementing HSRP and Implementing VRRP sections of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information about HSRP and VRRP.
Data Plane
The data plane contains the logical group of "customer" applications: traffic generated by hosts, clients, servers; and applications that are sourced from and destined to other similar devices supported by the network. Data plane traffic is typically forwarded in the fast path, although certain exception packets can be punted, requiring CPU assistance. Within the context of security and because the data plane represents the highest traffic rates, it is critical that the data plane be secured to prevent exception packets from punting to the CPU and impacting the control and management planes.
Within the data plane, there are many features and configuration options that can help secure traffic. The following sections detail these features and options that can be implemented to further secure the network.
Filter Transit Traffic with ACLs
Administrators can control what traffic transits the network by using transit access control lists (tACLs). The tACLs contrast with the infrastructure ACLs that seek to filter traffic that is destined to the network device. The filtering provided by tACLs is beneficial when administrators want to filter traffic to a particular group of devices or traffic that is transiting the network.
tACLs are also an appropriate mechanism with which to implement static anti-spoofing protections. Refer to the Anti-Spoofing Protections section of this guide and the Implementing Access Lists and Prefix Lists section of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information.
IPv4 and IPv6 ACL with Counters
In Cisco IOS XR Software, ACL counters are maintained in hardware and software. Hardware counters are used for packet-filtering applications, such as when an access group is applied to an interface. Software counters are used by all the applications and mainly pertain to software packet processing.
To display the hardware counters for an access group, use the following command:
show {ipv4 | ipv6} access-listsTo clear the hardware counters, use the following command:
clear {ipv4 | ipv6} access-listHardware counters are disabled by default for IPv4 ACLs. To enable hardware counting, use the following command:
ipv4 access-group access-list-name {in | out} [hw-count] Hardware counters are enabled only on the specified interface.
Software counters are updated only for the packets that are software processes—for example, exception packets that are punted to the LC CPU for processing, or an ACL that is used by routing protocols. This information also applies to IPv6 ACLs with one exception: hardware counting is always enabled for IPv6. Note that the hw-count option does not exist in the IPv6 access-group mode .
IPv6 ACL
In Cisco IOS XR Software, every IPv6 ACL has the following implicit statements as its last match entries:
permit icmp any any nd-na permit icmp any any nd-ns deny ipv6 any any
The first two match conditions allow for ICMPv6 neighbor discovery. The IPv6 neighbor discovery process makes use of the IPv6 network layer service; therefore, by default, IPv6 ACLs implicitly allow IPv6 neighbor discovery packets to be sent and received on an interface.
An IPv6 ACL must contain at least one entry for the implicit deny ipv6 any statement to take effect.
ICMP Packet Filtering
Certain ICMP message types are commonly used by network troubleshooting tools such as ping and trace route, as well as by Path MTU Discovery. There are other legitimate uses for ICMP. However, not all ICMP types are required for proper network operation, and ICMP flooding attacks can cause high CPU usage. It is necessary to have mechanisms that control which ICMP packet types are permitted to enter the network.
Cisco IOS XR Software provides ACL functionality to specifically filter ICMP messages by name or type and code. The following example shows IPv4 ACL allowing ICMP from trusted networks and permitting necessary ICMP type packets from other resources, while blocking all other ICMP type packets from other sources:
Ipv4 access-list ACL-TRANSIT-IN !!--- Permit ICMP packets from trusted networks only ! 10 permit icmp host x.x.x.x!!--- Permit other ICMP traffic, if required, for proper Customer ! !!--- network operations ! 20 permit icmp any any echo 30 permit icmp any any echo-reply 40 permit icmp any any ttl-exceeded 50 permit icmp any any unreachable 60 permit icmp any any packet-too-big !!--- Deny all other IP traffic to any network device ! 70 deny icmp any any
In the preceding example, 20 and 30 are for ping, 40 and 50 are for trace route, and 60 is for PMTUD.
Note: If the destination IP address happens to be receive, the preceding example is still relevant because Cisco IOS XR Software has LPTS to control packets that reach the CPU. As a result, there is a minimum risk of exhausting CPU resources. To prevent this risk on Cisco IOS devices, administrators should configure Control Plane Policing (CoPP).
Filter IPv4 Traffic with Remote Triggered Black Hole Filtering
Remote Triggered Black Hole Filtering uses the CEF process (fast path) to drop undesirable IPv4 traffic before it is forwarded to the network.
The following example provides a basic remote triggered black hole filtering configuration on an edge router:
community-set ddos-comm 1:666 end-set ! route-policy RTBH-drop if community matches-any ddos-comm then set next-hop 192.0.2.1 set local-preference 200 endif end-policy ! router static address-family ipv4 unicast 192.0.2.1/32 Null0 ! router bgp 65000 neighbor 191.0.0.2 remote-as 65188 route-policy RTBH-drop in !
Note: In Cisco IOS XR Software, the Null 0 interface always exists and it never sends ICMP unreachable messages.
Filter IPv6 Traffic with Remote Triggered Black Hole Filtering
Remote Triggered Black Hole Filtering uses the CEF process (fast path) to drop undesirable IPv6 traffic before forwarding that traffic to the network.
The following example provides a basic remote triggered black hole filtering configuration on an Edge Router:
community-set ddos-comm 1:666 end-set ! route-policy RTBH-drop-v6 if community matches-any ddos-comm then set next-hop 2001:db8:bad::1 endif end-policy ! router static address-family ipv6 unicast 2001:db8:bad::/48 Null0 ! router bgp 65000 neighbor 191.0.0.2 address-family ipv6 unicast route-policy RTBH-drop-v6 in !
Anti-Spoofing Protections
Many attacks utilize source IP address spoofing to conceal the true source of an attack and hinder accurate trace back or to defeat access control lists (ACLs). Cisco IOS XR Software provides the Unicast Reverse Path Forwarding (RPF) feature to deter attacks that rely on source IP address spoofing. In addition, ACLs and null routing are often deployed to augment spoofing protection.
Unicast Reverse Path Forwarding
Unicast Reverse Path Forwarding (Unicast RPF) provides source network verification and can reduce spoofed attacks from networks that are not under direct administrative control.
Unicast RPF enables a device to verify that the source address of a forwarded packet can be reached through the interface that received the packet. Administrators must not rely on Unicast RPF as the only protection against spoofing. Spoofed packets can enter the network through a Unicast RPF-enabled interface if an appropriate return route to the source IP address exists. Unicast RPF relies on the Cisco Express Forwarding feature on each device and is configured on a per-interface basis.
Unicast RPF can be configured in either of the following modes:
- Strict mode Unicast RPF: Strict mode checks the forwarding information base (FIB) to match the source address with the next hop adjacency. If the source address of the incoming packet does not match the next hop entry in the FIB for this address, the packet is dropped.
- Loose mode Unicast RPF: Loose mode searches for the source address of a packet in the FIB table. If the address exists and matches a real and valid forwarding entry (not necessarily pointing to the ingress interface on which the packet was received), then the packet is further processed, otherwise it is dropped.
In cases where there is asymmetric routing, loose mode is preferred because strict mode is known to drop packets. During configuration of the ipv4 verify interface configuration command, the keyword any configures loose mode, while the keyword rx configures strict mode.
The following example illustrates configuration of the ipv4 verify command:
interfaceipv4 verify unicast source reachable-via rx
or
ipv4 verify unicast source reachable-via any ! ipv6 verify unicast source reachable-via rx
or
ipv6 verify unicast source reachable-via any !
Reference the Implementing Cisco Express Forwarding section of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information.
Anti-Spoofing ACLs
Manually configured ACLs can provide static anti-spoofing protection against attacks that use known unused and untrusted address space. Commonly, these anti-spoofing ACLs are applied to ingress traffic at the network boundaries as a component of a larger ACL. Anti-spoofing ACLs require regular monitoring because they can frequently change. Spoofing can be minimized in traffic that originates from the local network by applying outbound ACLs that limit the traffic to valid local addresses.
The following example demonstrates how ACLs can be used to limit IP spoofing. The ACL is applied inbound on the desired interface. The access control entries (ACEs) that make up this ACL are not comprehensive. If administrators configure this type of ACL, they are advised to seek a conclusive, up to date reference.
ipv4 access-list ACL-ANTISPOOF-IN 10 deny ipv4 10.0.0.0 0.255.255.255 any 20 deny ipv4 192.168.0.0 0.0.255.255 any 30 permit ipv4 any any ! interfaceipv4 access-group ACL-ANTISPOOF-IN ingress !
See the IANA IPv4 Address Space Registry for a list of unallocated Internet addresses.
See the Implementing Access Lists and Prefix Lists section of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information about ACL configuration.
Limit CPU Impact of Data Plane Traffic
The primary purpose of routers is to forward packets and frames through the device to final destinations. These packets, which transit the devices deployed throughout the network, can impact the CPU operations of a device. The data plane, which consists of traffic transiting the network device, should be secured to ensure the operation of the management and control planes. If transit traffic can cause a device to process switch traffic, the control plane of a device can be affected, which may lead to an operational disruption.
Features and Traffic Types that Impact the RP and LC CPU
Although not exhaustive, the following list includes the types of data plane traffic that require special CPU processing and that are process-switched by the RP CPU or LC CPU:
- ACL Logging: ACL logging traffic consists of any packets that are generated due to a match (permit or deny) of an ACE on which the log keyword is used.
- IPv4 Options: Any IPv4 packets with options included must be processed by the CPU. Some may be processed by the LC CPU, while others may require full RP CPU support.
- Fragmentation: Any IPv4 packet that requires fragmentation must be passed to the LC CPU for processing.
- Time-to-Live (TTL) Expiry: Packets which have a TTL value less than or equal to 1 require ICMP Time Exceeded (ICMP Type 11, Code 0) messages to be sent, resulting in CPU processing. In most cases, processing is done by the LC CPU.
- ICMP Unreachable: IPv4 and IPv6 packets that require the generation of ICMP unreachable messages due to routing, MTU, or filtering are handled by either the LC CPU or the RP CPU.
- Traffic Requiring an ARP Request: Destinations for which an ARP entry does not exist (CEF glean) require processing by the CPU.
- Non-IPv4/IPv6 Traffic: All non-IPv4 and IPv6 traffic is processed by the CPU.
Some functions provided by the Cisco IOS ip options [ignore | drop] command and TTL ACL are not directly available in Cisco IOS XR Software; however, most ip options packets are rate-limited at the microcode level automatically in Cisco IOS XR software. Many packets can be managed by LPTS, which provides a function similar to the Cisco IOS Software Distributed Mode CoPP (dCoPP) capability. In Cisco IOS XR Software, LPTS is an automatic feature and does not require user configuration. The LPTS feature manages all traffic that would be handled by any CPU (line card of router processor) on a per-line card basis only (aggregate level rate-limiting at the RP level is not supported).
The policing values used within LPTS are not user-configurable until Cisco IOS XR Software Release 3.6. For Cisco IOS XR Software Releases versions 3.6 and later, the LPTS feature policing rates are user-configurable.
Traffic Identification and Traceback
At times, administrators must quickly identify and trace back network traffic, especially during an incident response or poor network performance. NetFlow and Classification ACLs are the two primary methods to accomplish traffic identification and traceback when using Cisco IOS XR Software. NetFlow can provide visibility into all traffic on the network. Additionally, NetFlow can be implemented with collectors that can provide long term trending and automated analysis. Classification ACLs are a component of ACLs. Preplanning is required to identify specific traffic and manual intervention during analysis. The following sections provide a brief overview of traffic identification and traceback.
Identify Anomalous Activity by Using NetFlow
NetFlow identifies anomalous and security-related network activity by tracking network flows. NetFlow data can be viewed and analyzed by way of the command line interface (CLI), or NetFlow records can be exported to a commercial or freeware NetFlow collector for aggregation and analysis. NetFlow collectors, through long-term trending, can provide network behavior and usage analysis. NetFlow functions by performing analysis on specific attributes within IP packets and creating flows. NetFlow version 5 is the most commonly used version; however, version 9 is more extensible. NetFlow flow records can be created using sampled traffic data in high-volume environments.
Administrators are advised to consider the following specifications when configuring NetFlow in Cisco IOS XR Software:
- Configure a source interface, otherwise, the exporter will remain in a disabled state
- Configure a valid record map name for every flow monitor map
- Only sampled NetFlow is supported (no static)
- Only the IPv4 raw record format is supported
- Up to 1 million cache entries are supported per flow monitor
- The DSCP value can be set for export packets to avoid policy drops
- Export only NetFlow version 9 record format over UDP
Administrators are advised not to use the management interface to export NetFlow records. Exporting NetFlow records using the management interface is not supported on the Cisco IOS XR 12000 Series Router and is not efficient on the Cisco CRS-1 Series routers.
Cisco IOS XR Software Release 3.5/3.6/3.7 supports the NetFlow collection of MPLS packets. It also supports the NetFlow collection of MPLS packets carrying IPv4, IPv6, or both IPv4 and IPv6 payloads. MPLS IPv6 is not supported on the Cisco IOS XR 12000 Series routers.
The sampler map, a subfeature of NetFlow, specifies the rate at which packets (one out of n packets from the flow) are sampled. On high bandwidth interfaces, applying NetFlow processing to every single packet can result in significant CPU utilization. Sampler map configuration is typically used for such high-speed interfaces. If NetFlow is applied in both directions, the flow record packets are policed at the rate of 35,000 packets per second per direction on the Cisco CRS-1, and 25,000 packets per second per direction on the Cisco IOS XR 12000 Series Router. If NetFlow is applied in one direction only, then the flow record packets are policed at the rate of 70,000 packets per second per direction on the Cisco CRS-1, and 50,000 packets per second per direction on the Cisco IOS XR 12000 Series Router. On CRS-1, MSC/B line cards have a higher policer rate than MSC/A.
The following is an example of NetFlow cache summary output from the CLI:
#show flow monitor PE3_monitor cache summary location 0/13/cPU0 Cache summary for Flow Monitor PE3_monitor: Cache size: 65535 Current entries: 627 High Watermark: 62258 Flows added: 3989508 Flows not added: 0 Ager Polls: 95895 • Active timeout 409 • Inactive timeout 3988471 • TCP FIN flag 1 • Watermark aged 0 • Emergency aged 0 • Counter wrap aged 0 • Total 3988881 Periodic export: • Counter wrap 0 • TCP FIN flag 0 Flows exported 3988881
See Configuring NetFlow on Cisco IOS XR Software for more information about NetFlow capabilities.
Identify Traffic by Using Classification ACLs
Classification ACLs provide visibility into the traffic that traverses an interface. Classification ACLs do not alter the security policy of a network and are typically constructed to classify individual protocols, source addresses, or destinations. For example, an ACE that permits all traffic could be separated into specific protocols or ports. This more granular classification of traffic into specific ACEs can help provide an understanding of the network traffic because each traffic category has its own hit counter. An administrator may also separate the implicit deny at the end of an ACL into granular ACEs to help identify the types of denied traffic.
An administrator can expedite an incident response by using classification ACLs with the show access-list and clear ip access-list counters EXEC commands.
The following example illustrates the configuration of a classification ACL to identify fragment traffic prior to a default deny:
ipv4 access-list permit-frag 10 permit tcp any any fragments log 20 permit udp any any fragments 30 permit icmp any any fragments 40 permit ipv4 any any fragments 45 permit icmp any any 50 permit ipv4 any any
To identify the traffic that uses a classification ACL, administrators should use the show access-list acl-name EXEC command. The ACL counters can be cleared by using the clear access-list ipv4 acl-name EXEC command.
# show access-lists permit-frag ipv4 access-list permit-frag 10 permit tcp any any fragments log (1386 matches) 20 permit udp any any fragments 30 permit icmp any any fragments 40 permit ipv4 any any fragments 45 permit icmp any any 50 permit ipv4 any any (11002 matches) !
See the Implementing Access Lists and Prefix Lists section of the Cisco IOS XR IP Addresses and Services Configuration Guide for more information about enabling logging capabilities within ACLs.
Conclusion
This document provides a broad overview of the methods that can be used to secure a Cisco IOS XR system device. Securing the devices increases the overall security of the networks that administrators manage. Administrators can increase the protection of the management, control, and data planes of Cisco IOS XR devices in their network by following the recommendations provided in this document.
Glossary
The following list provides expansions for acronyms and initialisms used in this document:
ARP: Address Resolution Protocol
ACE: Access Control Entries
AAA: Authentication, Authorization, and Accounting
AS: Autonomous System
Aux: Auxiliary Port
BGP: Border Gateway Protocol
CLI: Command Line Interface
CPU: Central Processing Unit
Cisco DP: Cisco Discovery Protocol
DF: Data Fragmentation
DNS: Domain Name Service
DoS: Denial of Service
DRP: Distributed Router Processor
dSC: Designated System Controller
eBGP: external Border Gateway Protocol
FIB: Forwarding Information Base
FTP: File Transfer Protocol
FHRPs: First Hop Redundancy Protocols
HSRP: Hot Standby Router Protocol
HTTP: Hypertext Transfer Protocol
ICMP: Internet Control Message Protocol
IGP: Interior Gateway Protocol
IPSec encryption: IP Security encryption
IS-IS: Intermediate System to Intermediate System
Ksh: korn shell
LDP: Label Distribution Protocol
LC: Line Card
LPTS: Local Packet Transport Services
MAC: Media Access Control
MPLS: Multiprotocol Label Switching
MPLS TE: Multiprotocol Label Switching Traffic Engineering
NTP: Network Time Protocol
OPIE: One-time Passwords In Everything
OSPF: Open Shortest Path First
PPP: Point to Point Protocol
PSIRT: Cisco Product Security Incident Response Team
Path MTU: Path Maximum Transmission Unit
Proxy ARP: Proxy Address Resolution Protocol
Qnet: QSSL Network Manager
RFC: Request for Comments
ROMMON: Read Only Memory Monitor
RP: Router Processor
RPL: Routing Policy Language
SC: Shelf Controller
SCP: Secure Copy Protocol
SDR: Secure Domain Router
SFTP: Secure FTP
SSH: Secure Shell Protocol
SNMP: Simple Network Management Protocol
Syslog: System Log
TACACS+: Terminal Access Controller Access-Control System
Telnet: Telecommunication Network
TFTP: Trivial File Transfer Protocol
TTL: Time-to-Live
Unicast RPF: Unicast Reverse Path Forwarding
VRRP: Virtual Router Redundancy Protocol
VRF: Virtual Routing and Forwarding
Acknowledgments
Chao Yu (ychao@cisco.com)
Technical Leader, Cisco IOS XR Team
Laura Kuiper (kuiperl@cisco.com)
Security Consulting Engineer
Tomasz Kalkowski, PhD (tkalkows@cisco.com)
Technical Leader, Cisco IOS XR Team
References
Cisco Guide to Harden Cisco IOS Devices
//www.cisco.com/en/US/tech/tk648/tk361/technologies_tech_note09186a0080120f48.shtml
Cisco IOS NetFlow
//www.cisco.com/en/US/products/ps6601/products_ios_protocol_group_home.html
Cisco IOS XR IP Addresses and Services Command Reference for Cisco CRS Routers
//www.cisco.com/c/en/us/td/docs/routers/crs/software/crs_r4-2/addr_serv/command/reference/b_ipaddr_cr42crs.html
Cisco ASR 9000 Series Aggregation Services Router IP Addresses and Services Configuration Guide, Release 5.3.x
//www.cisco.com/c/en/us/td/docs/routers/asr9000/software/asr9k_r5-3/addr-serv/configuration/guide/b-ipaddr-cg53asr9k.html
Cisco IOS XR Routing Configuration Guide for the Cisco CRS Router, Release 5.3.x
//www.cisco.com/c/en/us/td/docs/routers/crs/software/crs_r5-3/routing/configuration/guide/b_routing_cg53xcrs.html
Cisco IOS XR System Management Configuration Guide for the Cisco CRS Router, Release 5.3.x
//www.cisco.com/c/en/us/td/docs/routers/crs/software/crs_r5-3/sysman/configuration/guide/b-sysman-cg-53xcrs.html
Cisco ASR 9000 Series Aggregation Services Router System Monitoring Configuration Guide, Release 5.3.x
//www.cisco.com/c/en/us/td/docs/routers/asr9000/software/asr9k_r5-3/system-monitoring/configuration/guide/b_sysmon_cg53xasr9k.htmll
Cisco ASR 9000 Series Aggregation Services Router System Security Configuration Guide, Release 5.3.x
//www.cisco.com/c/en/us/td/docs/routers/asr9000/software/asr9k_r5-3/security/configuration/guide/b-syssec-cg53x-asr9k.html
Cisco Security Advisories and Alerts
//tools.cisco.com/security/center/publicationListing.x
Cisco Security Vulnerability Policy
//www.cisco.com/c/en/us/about/security-center/security-vulnerability-policy.html
RFC 2205: Resource ReSerVation Protocol (RSVP)
http://www.ietf.org/rfc/rfc2205.txt
RFC 2385: Protection of BGP Sessions via the TCP MD5 Signature Option
http://www.ietf.org/rfc/rfc2385.txt
RFC 2747: RSVP Cryptographic Authentication
http://www.ietf.org/rfc/rfc2747
RFC 3682: The Generalized TTL Security Mechanism (GTSM)
http://www.ietf.org/rfc/rfc3682.txt
RFC 3882: Configuring BGP to Block Denial-of-Service Attacks
http://www.ietf.org/rfc/rfc3882.txt
Risk Triage for Security Vulnerability Announcements
//www.cisco.com/c/en/us/about/security-center/vulnerability-risk-triage.html
TACACS+ and RADIUS Comparison
//www.cisco.com/en/US/tech/tk59/technologies_tech_note09186a0080094e99.shtml
The BGP TTL Security Hack (BTSH)
http://smakd.potaroo.net/ietf/idref/draft-gill-btsh/index.html
This document is part of the Cisco Security portal. Cisco provides the official information contained on the Cisco Security portal in English only.
This document is provided on an “as is” basis and does not imply any kind of guarantee or warranty, including the warranties of merchantability or fitness for a particular use. Your use of the information in the document or materials linked from the document is at your own risk. Cisco reserves the right to change or update this document without notice at any time.