Customer Premises Equipment
Service provider networks need to connect to many orders of magnitude more remote devices than enterprise networks do. Whereas a typical enterprise might have fewer than a hundred remote sites, a service provider could have thousands if not millions of subscribers. That means that scalability is of the utmost importance, and the reliability of the network must be maintained. IPv6 can uniquely support the addressing requirements for these types of networks.
No matter what type of physical medium the customer connection uses (xDSL, Cable/HFC, Fiber to the Home [FTTH], wireless), networking equipment called customer premises equipment (CPE) terminates service at the customer site. This equipment terminates the type of service provided by the Network Access Provider (NAP) and contains the Layer 3 address provided by the Network Service Provider (NSP). Each type of broadband access has its own way of connecting the customer to the NSP's routed infrastructure. Broadband access connections can use a direct connection or some form of tunneled protocol such as PPP or L2TP to connect the CPE to the Internet.
The service provider must secure its own network infrastructure when providing IPv6 services. Figure 3-5 shows several broadband access provider network topologies.
Figure 3-5 Broadband-Access Provider Topologies
Regardless of whether DSL Access Multiplexer (DSLAM)/Broadband Remote Access Server (BRAS) or Cable Modem Termination System (CMTS) devices are used, the edge router (ER) is the device that connects the customer's connection to the IPv6 Internet. These devices, and particularly the ER, must be hardened from a security perspective. The NAP and NSP should be able to keep track of which user's CPE has been allocated to which address or address block. This is typically done with RADIUS servers that authenticate the user's connection before allowing them on the network.
CPE for small offices/home offices (SOHO) or residential broadband access needs to be easy to configure, administer, and secure. This is a requirement because the users are not necessarily knowledgeable about IPv6 or even networking. If the device has advanced settings that give more control over the IPv6 connectivity and security, the default settings should be set to make it easy and yet secure. Otherwise, the service provider can have substantial support calls from customers to help them configure their CPE devices. Even though the end user should be concerned about the security of his connection to the Internet, he does not necessarily need to be bogged down in the details. Service providers must consider the customer's security and their end-user experience when selecting CPE devices on their behalf.
These home-user-grade products are the simplest form of routers. They have a single default route, and they provide DHCP services to the computers on their wired or wireless LAN. They gain a single public IP address and perform Port Address Translation (PAT) for the private addresses used behind the public IP. The security functions they provide involve simply preventing inbound connections from being made. Simple dual-stack residential devices have similar features for both protocols. However, for the IPv6 protocol, PAT is not necessary because global unicast addresses are used for the CPE device's external and internal interfaces.
Residential and SOHO security devices should perform some of the same functions as commercial-grade firewall products. Residential gateways need to be able to statefully permit outbound connections and only allow inbound packets that result from an outbound connection. If a CPE product does allow more advanced configuration of the firewall policy, the default settings should be used to prevent Internet traffic from reaching the internal LAN.
Consumer-grade CPE best practices include the following:
- Do not forward packets that have a multicast source address.
- Block packets destined for multicast destinations in the outbound direction.
- Do not allow RH0 packets inbound or outbound.
- Block packets sourced from Unique Local Address (ULA) space (that is, FC00::/7).
- Block other bogon addresses from entering and leaving the interior LAN (difficult to do because this list changes several times each year).
- Block packets that are not sourced from the global unicast prefix assigned on the LAN interface. This prevents spoofed packets from leaving the user LAN.
- Deny packets sourced from the internal LAN prefix from coming in the external interface. This prevents spoofed packets from entering from the Internet.
IPv6-capable CPE routers should also prevent Teredo tunnels from forming from internal clients to Teredo servers. Teredo is only a transition mechanism for IPv6-capable hosts behind IPv4 NATs. Teredo is not used if the client has native IPv6 connectivity through the CPE router to an IPv6-capable service provider. The risk is that Teredo tunnels can be used as a back door into the client computer. However, preventing Teredo tunnels from being established can be difficult to accomplish. More information on this subject appears in Chapter 10.
If consumer-grade CPE devices are constructed with IPv6 security measures enabled by default, the customer's Internet connection will be more secure. The end user does not have to be so worried about these details, and the device can provide the required security features right out of the box.
More advanced users, like those reading this book, might want to have a more sophisticated device at their homes. For more advanced users that require more power, CPE devices such as the Cisco ASA5505, the Cisco 871, or the newer Cisco 880/860 routers perform nicely. These devices have full IPv6 capabilities and the ability to filter IPv6 packets based on a wide variety of header fields and extension headers. These devices might not have these default security settings. Therefore, the more advanced users need to be able to configure these same settings to secure their own Internet connections.