3. MARCO TEÓRICO
3.3 TRACTOS DE LA MÉDULA ESPINAL
PE routers associate physical (or logical) interfaces to customer VPNs using VRFs. VRFs are statically assigned to interfaces and cannot be modified without PE router reconfiguration. Techniques are available for the PE router to dynamically select a VRF (or VPN) based on the IP source address of traffic received from the CE router or through policy-based routing (PBR); however, these are not generally recommended for site-to-site MPLS VPN services, given the threat of spoofed IP attacks, and hence are not widely deployed. Using a static VRF configuration provides complete separation between VPNs, and between VPNs and the SP global IP routing table.
VPN customer packets cannot travel outside of the assigned VPN unless the SP VPN policies explicitly allow for it. Conversely, external packets cannot be injected inside the VPN unless explicitly allowed. RFC 3032 and IETF draft-ietf-mpls-icmp-07 specify the interaction between MPLS and ICMP, and allow for core (P) router generated ICMP messages to be sent to a source IP host within a customer VPN as required, for example, ICMP Time Exceeded (Type 11) and ICMP Destination Unreachable (Type 3) due to “Fragmentation Needed and Don’t Fragment was Set” (Code 4).
With the exception of carrier-supporting-carrier configurations detailed in the “Threats Against the Inter-Provider Edge” section below, all MPLS labeled packets received on a VRF interface or a native IP interface (for example, PE-CE) not enabled for MPLS will be discarded. This prevents an attacker from injecting unauthorized packets into the VPN through the use of spoofed MPLS labels. MPLS VPNs also prevent against external attempts to hijack customer VPN routes. If a malicious or compromised CE router advertises to a PE router an IPv4 route tagged with an illegal route target (RT), the PE router will strip the illegal RT off and only advertise the PE-configured export RT list when converted to a VPN prefix and advertised within M-iBGP. Similar attempts to hijack a VPN prefix through eBGP on a native IP external interface also do not pose a risk, because these BGP
prefixes use separate subsequent address family identifiers (SAFI). In summary, only a misconfiguration or software vulnerability would allow illegal VPN packets or prefixes to leak into a VPN.
Although the PE router provides routing and address separation between VPNs, it is also reachable by its IP address within each configured VPN. This makes it vulnerable to internal IP attacks sourced from within a VPN. If an Ethernet switch is used to connect the PE router with CE routers, then the Layer 2 Ethernet threats may also apply. Further, given that a PE router aggregates many customers and VPNs, an attack against the PE within one VPN may adversely affect other VPN customers. This is due to the PE router sharing its resources, including CPU, memory, and internal (uplink) interface bandwidth, among the different customer VPNs. Hence, withstanding internal attacks and unauthorized access to the SP network infrastructure, the impact of collateral damage is the most significant threat against MPLS VPNs.
The risk of this threat may increase if the PE router also delivers Internet transit services. In this scenario, an Internet attack against the PE router (or an attached Internet transit customer) may trigger collateral damage within the PE router, thereby adversely affecting VPN customers attached to the same PE router. Collateral damage may cause packet loss, which may then trigger Layer 2 or Layer 3 protocol timeouts. In this event, affected interfaces and routing protocols may fail, resulting in loss of VPN connectivity. Hence, although an MPLS VPN assures routing and addressing separation between VPNs, and between VPNs and the SP global IP routing table, collateral damage remains a very real threat. Techniques to mitigate the risk of collateral damage are available and are reviewed in Part II. Additional threats against the PE router include: