Juniper Router RTR Security Technical Implementation Guide

Information flow control regulates where information is allowed to travel within a network and between interconnected networks. The flow of all network traffic must be monitored and controlled so it does not introduce any unacceptable risk to the network infrastructure or data. Information flow control policies and enforcement mechanisms are commonly employed by organizations to control the flow of information between designated sources and destinations (e.g., networks, individuals, and devices) within information systems. Enforcement occurs, for example, in boundary protection devices (e.g., gateways, routers, guards, encrypted tunnels, and firewalls) that employ rule sets or establish configuration settings that restrict information system services, provide a packet filtering capability based on header information, or provide a message filtering capability based on message content (e.g., implementing key word searches or using document characteristics).

A rogue router could send a fictitious routing update to convince a site's perimeter router to send traffic to an incorrect or even a rogue destination. This diverted traffic could be analyzed to learn confidential information about the site's network or used to disrupt the network's ability to communicate with other networks. This is known as a "traffic attraction attack" and is prevented by configuring neighbor router authentication for routing updates. This requirement applies to all IPv4 and IPv6 protocols that are used to exchange routing or packet forwarding information; this includes all Interior Gateway Protocols (such as OSPF, EIGRP, and IS-IS) and Exterior Gateway Protocols (such as BGP), MPLS-related protocols (such as LDP), and multicast-related protocols.

If the keys used for routing protocol authentication are guessed, the malicious user could create havoc within the network by advertising incorrect routes and redirecting traffic. Some routing protocols allow the use of key chains for authentication. A key chain is a set of keys that is used in succession, with each having a lifetime of no more than 180 days. Changing the keys frequently reduces the risk of them eventually being guessed. Keys cannot be used during time periods for which they are not activated. If a time period occurs during which no key is activated, neighbor authentication cannot occur, and therefore routing updates will fail. Therefore, ensure that for a given key chain, key activation times overlap to avoid any period of time during which no key is activated.

A rogue router could send a fictitious routing update to convince a site's perimeter router to send traffic to an incorrect or even a rogue destination. This diverted traffic could be analyzed to learn confidential information about the site's network or used to disrupt the network's ability to communicate with other networks. This is known as a "traffic attraction attack" and is prevented by configuring neighbor router authentication for routing updates. However, using clear-text authentication provides little benefit since an attacker can intercept traffic and view the authentication key. This would allow the attacker to use the authentication key in an attack. This requirement applies to all IPv4 and IPv6 protocols that are used to exchange routing or packet forwarding information; this includes all Interior Gateway Protocols (such as OSPF, EIGRP, and IS-IS) and Exterior Gateway Protocols (such as BGP), MPLS-related protocols (such as LDP), and multicast-related protocols.

A rogue router could send a fictitious routing update to convince a site's perimeter router to send traffic to an incorrect or even a rogue destination. This diverted traffic could be analyzed to learn confidential information about the site's network or used to disrupt the network's ability to communicate with other networks. This is known as a "traffic attraction attack" and is prevented by configuring neighbor router authentication for routing updates. However, using clear-text authentication provides little benefit since an attacker can intercept traffic and view the authentication key. This would allow the attacker to use the authentication key in an attack. Since MD5 is vulnerable to "birthday" attacks and may be compromised, routing protocol authentication must use FIPS 140-2 validated algorithms and modules to encrypt the authentication key. This requirement applies to all IPv4 and IPv6 protocols that are used to exchange routing or packet forwarding information; this includes all Interior Gateway Protocols (such as OSPF, EIGRP, and IS-IS) and Exterior Gateway Protocols (such as BGP), MPLS-related protocols (such as LDP), and multicast-related protocols.

The Route Processor (RP) is critical to all network operations because it is the component used to build all forwarding paths for the data plane via control plane processes. It is also instrumental with ongoing network management functions that keep the routers and links available for providing network services. Any disruption to the RP or the control and management planes can result in mission-critical network outages. A DoS attack targeting the RP can result in excessive CPU and memory utilization. To maintain network stability and RP security, the router must be able to handle specific control plane and management plane traffic that is destined to the RP. In the past, one method of filtering was to use ingress filters on forwarding interfaces to filter both forwarding path and receiving path traffic. However, this method does not scale well as the number of interfaces grows and the size of the ingress filters grows. Control plane policing increases the security of routers and multilayer switches by protecting the RP from unnecessary or malicious traffic. Filtering and rate limiting the traffic flow of control plane packets can be implemented to protect routers against reconnaissance and DoS attacks, allowing the control plane to maintain packet forwarding and protocol states despite an attack or heavy load on the router or multilayer switch.

Fragmented ICMP packets can be generated by hackers for DoS attacks such as Ping O' Death and Teardrop. It is imperative that all fragmented ICMP packets are dropped.

A gratuitous ARP is an ARP broadcast in which the source and destination MAC addresses are the same. It is used to inform the network about a host IP address. A spoofed gratuitous ARP message can cause network mapping information to be stored incorrectly, causing network malfunction.

The ICMP supports IP traffic by relaying information about paths, routes, and network conditions. Routers automatically send ICMP messages under a wide variety of conditions. Host unreachable ICMP messages are commonly used by attackers for network mapping and diagnosis.

The ICMP supports IP traffic by relaying information about paths, routes, and network conditions. Routers automatically send ICMP messages under a wide variety of conditions. Mask Reply ICMP messages are commonly used by attackers for network mapping and diagnosis.

The ICMP supports IP traffic by relaying information about paths, routes, and network conditions. Routers automatically send ICMP messages under a wide variety of conditions. Redirect ICMP messages are commonly used by attackers for network mapping and diagnosis.

Without establishing where events occurred, it is impossible to establish, correlate, and investigate the events leading up to an outage or attack. In order to compile an accurate risk assessment and provide forensic analysis, it is essential for security personnel to know where events occurred, such as router components, modules, device identifiers, node names, and functionality. Associating information about where the event occurred within the network provides a means of investigating an attack, recognizing resource utilization or capacity thresholds, or identifying an improperly configured router.

Without establishing the source of the event, it is impossible to establish, correlate, and investigate the events leading up to an outage or attack. In order to compile an accurate risk assessment and provide forensic analysis, security personnel need to know the source of the event. In addition to logging where events occur within the network, the audit records must also identify sources of events such as IP addresses, processes, and node or device names.

Information flow control regulates authorized information to travel within a network and between interconnected networks. Controlling the flow of network traffic is critical so it does not introduce any unacceptable risk to the network infrastructure or data. An example of a flow control restriction is blocking outside traffic claiming to be from within the organization. For most routers, internal information flow control is a product of system design.

Unrestricted traffic may contain malicious traffic that poses a threat to an enclave or to other connected networks. Additionally, unrestricted traffic may transit a network, which uses bandwidth and other resources. Traffic can be restricted directly by an access control list (ACL), which is a firewall function, or by Policy Routing. Policy Routing is a technique used to make routing decisions based on a number of different criteria other than just the destination network, including source or destination network, source or destination address, source or destination port, protocol, packet size, and packet classification. This overrides the router's normal routing procedures used to control the specific paths of network traffic. It is normally used for traffic engineering but can also be used to meet security requirements; for example, traffic that is not allowed can be routed to the Null0 or discard interface. Policy Routing can also be used to control which prefixes appear in the routing table. This requirement is intended to allow network administrators the flexibility to use whatever technique is most effective.

Packets with Bogon IP source addresses should never be allowed to traverse the IP core. Bogon IP networks are RFC1918 addresses or address blocks that have never been assigned by the IANA or have been reserved.

Vulnerability assessments must be reviewed by the System Administrator, and protocols must be approved by the Information Assurance (IA) staff before entering the enclave. Access control lists (ACLs) are the first line of defense in a layered security approach. They permit authorized packets and deny unauthorized packets based on port or service type. They enhance the posture of the network by not allowing packets to reach a potential target within the security domain. The lists provided are highly susceptible ports and services that should be blocked or limited as much as possible without adversely affecting customer requirements. Auditing packets attempting to penetrate the network but that are stopped by an ACL will allow network administrators to broaden their protective ring and more tightly define the scope of operation. If the perimeter is in a Deny-by-Default posture and what is allowed through the filter is in accordance with DoD Instruction 8551.1, and if the permit rule is explicitly defined with explicit ports and protocols allowed, then all requirements related to PPS being blocked would be satisfied.

Access lists are used to separate data traffic into that which it will route (permitted packets) and that which it will not route (denied packets). Secure configuration of routers makes use of access lists for restricting access to services on the router itself as well as for filtering traffic passing through the router. Inbound versus Outbound: It should be noted that some operating systems default access lists are applied to the outbound queue. The more secure solution is to apply the access list to the inbound queue for three reasons: - The router can protect itself before damage is inflicted. - The input port is still known and can be filtered upon. - It is more efficient to filter packets before routing them.

Access lists are used to separate data traffic into that which it will route (permitted packets) and that which it will not route (denied packets). Secure configuration of routers makes use of access lists for restricting access to services on the router itself as well as for filtering traffic passing through the router. Inbound versus Outbound: It should be noted that some operating systems default access lists are applied to the outbound queue. The more secure solution is to apply the access list to the inbound queue for three reasons: - The router can protect itself before damage is inflicted. - The input port is still known and can be filtered upon. - It is more efficient to filter packets before routing them.

Packets with IP options are not fast switched and henceforth must be punted to the router processor. Hackers who initiate denial-of-service (DoS) attacks on routers commonly send large streams of packets with IP options. Dropping the packets with IP options reduces the load of IP options packets on the router. The end result is a reduction in the effects of the DoS attack on the router and on downstream routers.

When Proxy ARP is enabled on a Cisco router, it allows that router to extend the network (at Layer 2) across multiple interfaces (LAN segments). Because proxy ARP allows hosts from different LAN segments to look like they are on the same segment, proxy ARP is only safe when used between trusted LAN segments. Attackers can leverage the trusting nature of proxy ARP by spoofing a trusted host and then intercepting packets. Proxy ARP should always be disabled on router interfaces that do not require it, unless the router is being used as a LAN bridge.

For in-band management, the management network must have its own subnet in order to enforce control and access boundaries provided by Layer 3 network nodes, such as routers and firewalls. Management traffic between the managed network elements and the management network is routed via the same links and nodes as that used for production or operational traffic. Safeguards must be implemented to ensure that the management traffic does not leak past the perimeter of the managed network.

Using dedicated paths, the OOBM backbone connects the OOBM gateway routers located at the edge of the managed network and at the NOC. Dedicated links can be deployed using provisioned circuits or MPLS Layer 2 and Layer 3 VPN services or implementing a secured path with gateway-to-gateway IPsec tunnels. The tunnel mode ensures that the management traffic will be logically separated from any other traffic traversing the same path.

The OOBM network is an IP network used exclusively for the transport of OAM&P data from the network being managed to the OSS components located at the NOC. Its design provides connectivity to each managed network device, enabling network management traffic to flow between the managed network elements and the NOC. This allows the use of paths separate from those used by the managed network.

If the gateway router is not a dedicated device for the OOBM network, implementation of several safeguards for containment of management and production traffic boundaries must occur. Since the managed and management network are separate routing domains, configuration of separate Interior Gateway Protocol routing instances is critical on the router to segregate traffic from each network.

If the gateway router is not a dedicated device for the OOBM network, several safeguards must be implemented for containment of management and production traffic boundaries; otherwise, it is possible that management traffic will not be separated from production traffic. Since the managed network and the management network are separate routing domains, separate Interior Gateway Protocol routing instances must be configured on the router, one for the managed network and one for the OOBM network. In addition, the routes from the two domains must not be redistributed to each other.

If the gateway router is not a dedicated device for the OOBM network, several safeguards must be implemented for containment of management and production traffic boundaries. It is imperative that hosts from the managed network are not able to access the OOBM gateway router.

The OOBM access switch will connect to the management interface of the managed network elements. The management interface can be a true OOBM interface or a standard interface functioning as the management interface. In either case, the management interface of the managed network element will be directly connected to the OOBM network. An OOBM interface does not forward transit traffic, thereby providing complete separation of production and management traffic. Since all management traffic is immediately forwarded into the management network, it is not exposed to possible tampering. The separation also ensures that congestion or failures in the managed network do not affect the management of the device. If the device does not have an OOBM port, the interface functioning as the management interface must be configured so that management traffic does not leak into the managed network and that production traffic does not leak into the management network.

When the production network is managed in-band, the management network could be housed at a NOC that is located remotely at single or multiple interconnected sites. NOC interconnectivity, as well as connectivity between the NOC and the managed network, must be enabled using IPsec tunnels to provide the separation and integrity of the managed traffic.

If the same keys are used between eBGP neighbors, the chance of a hacker compromising any of the BGP sessions increases. It is possible that a malicious user exists in one autonomous system who would know the key used for the eBGP session. This user would then be able to hijack BGP sessions with other trusted neighbors.

Accepting route advertisements for Bogon prefixes can result in the local autonomous system (AS) becoming a transit for malicious traffic as it will in turn advertise these prefixes to neighbor autonomous systems.

Accepting route advertisements belonging to the local AS can result in traffic looping or being black holed, or at a minimum using a non-optimized path.

As a best practice, a service provider should only accept customer prefixes that have been assigned to that customer and any peering autonomous systems. A multi-homed customer with BGP speaking routers connected to the Internet or other external networks could be breached and used to launch a prefix de-aggregation attack. Without ingress route filtering of customers, the effectiveness of such an attack could impact the entire IP core and its customers.

Advertisement of routes by an autonomous system for networks that do not belong to any of its customers pulls traffic away from the authorized network. This causes a denial of service (DoS) on the network that allocated the block of addresses and may cause a DoS on the network that is inadvertently advertising it as the originator. It is also possible that a misconfigured or compromised router within the GIG IP core could redistribute IGP routes into BGP, thereby leaking internal routes.

Outbound route advertisements belonging to the core can result in traffic either looping or being black holed, or at a minimum, using a non-optimized path.

The effects of prefix de-aggregation can degrade router performance due to the size of routing tables and also result in black-holing legitimate traffic. Initiated by an attacker or a misconfigured router, prefix de-aggregation occurs when the announcement of a large prefix is fragmented into a collection of smaller prefix announcements. In 1997, misconfigured routers in the Florida Internet Exchange network (AS7007) de-aggregated every prefix in their routing table and started advertising the first /24 block of each of these prefixes as their own. Faced with this additional burden, the internal routers became overloaded and crashed repeatedly. This caused prefixes advertised by these routers to disappear from routing tables and reappear when the routers came back online. As the routers came back after crashing, they were flooded with the routing table information by their neighbors. The flood of information would again overwhelm the routers and cause them to crash. This process of route flapping served to destabilize not only the surrounding network but also the entire Internet. Routers trying to reach those addresses would choose the smaller, more specific /24 blocks first. This caused backbone networks throughout North America and Europe to crash. Maximum prefix limits on peer connections combined with aggressive prefix-size filtering of customers' reachability advertisements will effectively mitigate the de-aggregation risk. BGP maximum prefix must be used on all eBGP routers to limit the number of prefixes that it should receive from a particular neighbor, whether customer or peering AS. Consider each neighbor and how many routes they should be advertising and set a threshold slightly higher than the number expected.

The head end of the label-switched path (LSP), the label edge router (LER) will decrement the IP packet's time-to-live (TTL) value by one and then copy the value to the MPLS TTL field. At each label-switched router (LSR) hop, the MPLS TTL value is decremented by one. The MPLS router that pops the label (either the penultimate LSR or the egress LER) will copy the packet's MPLS TTL value to the IP TTL field and decrement it by one. This TTL propagation is the default behavior. Because the MPLS TTL is propagated from the IP TTL, a traceroute will list every hop in the path, be it routed or label switched, thereby exposing core nodes. With TTL propagation disabled, LER decrements the IP packet's TTL value by one and then places a value of 255 in the packet's MPLS TTL field, which is then decremented by one as the packet passes through each LSR in the MPLS core. Because the MPLS TTL never drops to zero, none of the LSP hops triggers an ICMP TTL exceeded message and consequently, these hops are not recorded in a traceroute. Hence, nodes within the MPLS core cannot be discovered by an attacker.

An RD provides uniqueness to the customer address spaces within the MPLS L3VPN infrastructure. The concept of the VPN-IPv4 and VPN-IPv6 address families consists of the RD prepended before the IP address. Hence, if the same IP prefix is used in several different L3VPNs, it is possible for BGP to carry several completely different routes for that prefix, one for each VPN. Since VPN-IPv4 addresses and IPv4 addresses are different address families, BGP never treats them as comparable addresses. The purpose of the RD is to create distinct routes for common IPv4 address prefixes. On any given PE router, a single RD can define a VRF in which the entire address space may be used independently, regardless of the makeup of other VPN address spaces. Hence, it is imperative that a unique RD is assigned to each L3VPN and that the proper RD is configured for each VRF.

LDP provides the signaling required for setting up and tearing down pseudowires (virtual circuits used to transport Layer 2 frames) across an MPLS IP core network. Using a targeted LDP session, each PE router advertises a virtual circuit label mapping that is used as part of the label stack imposed on the frames by the ingress PE router during packet forwarding. Authentication provides protection against spoofed TCP segments that can be introduced into the LDP sessions.

A traffic storm occurs when packets flood a VPLS bridge, creating excessive traffic and degrading network performance. Traffic storm control prevents VPLS bridge disruption by suppressing traffic when the number of packets reaches configured threshold levels. Traffic storm control monitors incoming traffic levels on a port and drops traffic when the number of packets reaches the configured threshold level during any one-second interval.

VPLS defines an architecture that delivers Ethernet multipoint services over an MPLS network. Customer Layer 2 frames are forwarded across the MPLS core via pseudowires using IEEE 802.1q Ethernet bridging principles. A pseudowire is a virtual bidirectional connection between two attachment circuits (virtual connections between PE and CE routers). A pseudowire contains two unidirectional label-switched paths (LSP). Each MAC forwarding table instance is interconnected using domain-specific LSPs, thereby maintaining privacy and logical separation between each VPLS domain. When a frame arrives on a bridge port (pseudowire or attachment circuit) and the source MAC address is unknown to the receiving PE router, the source MAC address is associated with the pseudowire or attachment circuit and the forwarding table is updated accordingly. Frames are forwarded to the appropriate pseudowire or attachment circuit according to the forwarding table entry for the destination MAC address. Ethernet frames sent to broadcast and unknown destination addresses must be flooded out to all interfaces for the bridge domain; hence, a PE router must replicate packets across both attachment circuits and pseudowires. A malicious attacker residing in a customer network could launch a source MAC address spoofing attack by flooding packets to a valid unicast destination, each with a different MAC source address. The PE router receiving this traffic would try to learn every new MAC address and would quickly run out of space for the VFI forwarding table. Older, valid MAC addresses would be removed from the table, and traffic sent to them would have to be flooded until the storm threshold limit is reached. Hence, it is essential that a limit is established to control the number of MAC addresses that will be learned and recorded into the forwarding table for each bridge domain.

The uRPF feature is a defense against spoofing and denial-of-service (DoS) attacks by verifying if the source address of any ingress packet is reachable. To mitigate attacks that rely on forged source addresses, all provider edge routers must enable uRPF loose mode to guarantee that all packets received from a CE router contain source addresses that are in the route table.

Packets with IP options are not fast switched and therefore must be punted to the router processor. Hackers who initiate denial-of-service (DoS) attacks on routers commonly send large streams of packets with IP options. Dropping the packets with IP options reduces the load of IP options packets on the router. The end result is a reduction in the effects of the DoS attack on the router and on downstream routers.

QoS enables DISA to offer value-added IP services in accordance with SLAs, ensuring that customer requirements can be met while providing a method to provision the edge and core to accommodate those requirements. The IP core will recognize and provide forwarding treatment of customer traffic according to the Differentiated Services Code Points (DSCP). Customers marking traffic within their DiffServ domain will be required to comply with the DSCP classification that has been approved by the DOD QoS Working Group. Non-compliance could enable a customer or even an attacker to rob bandwidth from other customers or mission-critical services.

DoS is a condition when a resource is not available for legitimate users. Packet flooding distributed denial-of-service (DDoS) attacks are referred to as volumetric attacks and have the objective of overloading a network or circuit to deny or seriously degrade performance, which denies access to the services that normally traverse the network or circuit. Volumetric attacks have become relatively easy to launch using readily available tools such as Low Orbit Ion Cannon or botnets. Measures to mitigate the effects of a successful volumetric attack must be taken to ensure that sufficient capacity is available for mission-critical traffic. Managing capacity may include, for example, establishing selected network usage priorities or quotas and enforcing them using rate limiting, Quality of Service (QoS), or other resource reservation control methods. These measures may also mitigate the effects of sudden decreases in network capacity that are the result of accidental or intentional physical damage to telecommunications facilities (such as cable cuts or weather-related outages).

If multicast traffic is forwarded beyond the intended boundary, it is possible that it can be intercepted by unauthorized or unintended personnel. Limiting where, within the network, a given multicast group's data is permitted to flow is an important first step in improving multicast security. A scope zone is an instance of a connected region of a given scope. Zones of the same scope cannot overlap while zones of a smaller scope will fit completely within a zone of a larger scope. For example, Admin-local scope is smaller than Site-local scope, so the administratively configured boundary fits within the bounds of a site. According to RFC 4007 IPv6 Scoped Address Architecture (section 5), scope zones are also required to be "convex from a routing perspective"; that is, packets routed within a zone must not pass through any links that are outside of the zone. This requirement forces each zone to be one contiguous island rather than a series of separate islands. As stated in the DoD IPv6 IA Guidance for MO3, "One should be able to identify all interfaces of a zone by drawing a closed loop on their network diagram, engulfing some routers and passing through some routers to include only some of their interfaces." Therefore, it is imperative that the network engineers have documented their multicast topology and thereby knows which interfaces are enabled for multicast. Once this is done, the zones can be scoped as required.

PIM is a routing protocol used to build multicast distribution trees for forwarding multicast traffic across the network infrastructure. PIM traffic must be limited to only known PIM neighbors by configuring and binding a PIM neighbor filter to those interfaces that have PIM enabled. If a PIM neighbor filter is not applied to those interfaces that have PIM enabled, unauthorized routers can join the PIM domain, discover and use the rendezvous points, and also advertise their rendezvous points into the domain. This can result in a denial of service by traffic flooding or result in the unauthorized transfer of data.

When a new source starts transmitting in a PIM Sparse Mode network, the DR will encapsulate the multicast packets into register messages and forward them to the RP using unicast. This process can be taxing on the CPU for both the DR and the RP if the source is running at a high data rate and there are many new sources starting at the same time. This scenario can potentially occur immediately after a network failover. The rate limit for the number of register messages should be set to a relatively low value based on the known number of multicast sources within the multicast domain.

Real-time multicast traffic can entail multiple large flows of data. Large unicast flows tend to be fairly isolated (i.e., someone doing a file download here or there), whereas multicast can have broader impact on bandwidth consumption, resulting in extreme network congestion. Hence, it is imperative that there is multicast admission control to restrict which multicast groups hosts are allowed to join via IGMP or MLD.

The current multicast paradigm can let any host join any multicast group at any time by sending an IGMP or MLD membership report to the DR. In a Protocol Independent Multicast (PIM) Sparse Mode network, the DR will send a PIM Join message for the group to the RP. Without any form of admission control, this can pose a security risk to the entire multicast domain - specifically the multicast routers along the shared tree from the DR to the RP that must maintain the mroute state information for each group join request. Hence, it is imperative that the DR is configured to limit the number of mroute state information that must be maintained to mitigate the risk of IGMP or MLD flooding.

ASM can have many sources for the same groups (many-to-many). For many receivers, the path via the RP may not be ideal compared with the shortest path from the source to the receiver. By default, the last-hop router will initiate a switch from the shared tree to a source-specific SPT to obtain lower latencies. This is accomplished by the last-hop router sending an (S, G) Protocol Independent Multicast (PIM) Join toward S (the source). When the last-hop router begins to receive traffic for the group from the source via the SPT, it will send a PIM Prune message to the RP for the (S, G). The RP will then send a Prune message toward the source. The SPT switchover becomes a scaling issue for large multicast topologies that have many receivers and many sources for many groups because (S, G) entries require more memory than (*, G). Hence, it is imperative to minimize the amount of (S, G) state to be maintained by increasing the threshold that determines when the SPT switchover occurs.

MSDP peering with customer network routers presents additional risks to the DISN Core, whether from a rogue or misconfigured MSDP-enabled router. To guard against an attack from malicious MSDP traffic, the receive path or interface filter for all MSDP-enabled RP routers must be configured to only accept MSDP packets from known MSDP peers.

MSDP peering with customer network routers presents additional risks to the core, whether from a rogue or misconfigured MSDP-enabled router. MSDP password authentication is used to validate each segment sent on the TCP connection between MSDP peers, protecting the MSDP session against the threat of spoofed packets being injected into the TCP connection stream.