The best practice before deploying a machine into a production environment or connecting your network to the Internet is to determine your organizational needs and how security can fit into the requirements as transparently as possible. Since the main goal of Red Hat Linux Security Guide is to explain how to secure Red Hat Linux operating system, a more detailed examination of hardware and physical network security is beyond the scope of this document. However, this chapter is a brief overview of establishing security policies with regard to hardware and physical networks. Important factors to consider are how computing needs and connectivity requirements fit into the overall security strategy. The following explains some of these factors in detail.
Connectivity is the method by which an administrator intends to connect disparate resources on a network. An administrator may use Ethernet (hubbed or switched CAT-5/RJ-45 cabling), token ring, 10-base-2 coaxial cable, or even cable-free (wireless, 802.11x) technologies. Depending on which medium an administrator chooses, certain media and network topologies require complementary technologies such as hubs, routers, switches, base stations, and access points. Determining a functional network architecture will allow an easier administrative process if security issues arise.
Computing involves more than just workstations running desktop software. Modern organizations require massive computational power and highly-available services, which can include mainframes, compute/server clusters, powerful workstations, and specialized appliances. With these organizational requirements, however, come increased susceptibility to hardware failure, natural disasters, and tampering or theft of equipment.
From these general considerations, administrators can get a better view of implementation. The design of a computing environment will then be based on both organizational need and security considerations — a true, "ground-up" implementation that places priority on both factors.
The foundation of a LAN is the topology, or network architecture. A topology is the physical and logical layout of a LAN in terms of resource provided, distance between nodes, and transmission medium. Depending upon the needs of the organization that the network will service, there are several choices available for network implementation. Each topology has its advantages and security issues that network architects should regard when designing their network layout.
As defined by the Institute of Electrical and Electronics Engineers (IEEE), there are three common topologies for physical connection of a LAN.
The Ring topology connects each node by exactly two connections. This creates a ring structure where each node is accessible to the other either directly by its two physically closest neighboring nodes and indirectly through the physical ring. Token Ring, FDDI, and SONET networks are connected in this fashion (with FDDI utilizing a dual-ring technique); however, there are no common Ethernet connections using this physical topology, so rings are not commonly deployed except in legacy or institutional settings with a large installed base of nodes (for example, a university).
The linear bus topology consists of nodes which connect to a terminated main linear cable (the backbone). The linear bus topology requires the least amount of cabling and networking equipment, making it the most cost-effective topology. However, the linear bus depends on the backbone being constantly available, making it a single point-of-failure if it has to be taken off-line or is severed. Linear bus topologies are commonly used in peer-to-peer LANs using co-axial (coax) cabling and 50-93 ohm terminators at both ends of the bus.
The Star topology incorporates a central point where nodes connect and through which communication is passed. This centerpoint, called a hub can be either broadcasted or switched. This topology does introduce a single point of failure in the centralized networking hardware that will connect the nodes. However, because of this centralization, networking issues that affect segments or the entire LAN itself is easily traceable to this one source.
In a broadcast network, a node will send a packet that traverses through every other node until the recipient accepts the packet. Every node in the network will conceivably receive this packet of data until the recipient processes the packet. In a broadcast network, all packets are sent in this manner.
In a switched network, packets are not broadcasted, but are processed in the switched hub which, in turn, will create a direct connection between the sending and recipient nodes using the unicast transmission principles. This eliminates the need to broadcast packets to each node, thus lowering traffic overhead.
The switched network also prevents packets from being intercepted by malicious nodes or users. In a broadcast network, since each node receives the packet en route to its destination, malicious users can set their Ethernet device to promiscuous mode and accept all packets regardless of whether or not the data is intended for them. Once in promiscuous mode, a sniffer application can be used to filter, analyze, and reconstruct packets for passwords, personal data, and more. Sophisticated sniffer applications will store such information in a text file and, perhaps, even send the information to an arbitrary source (for example, the malicious user's email address).
A switched network requires a network switch, a specialized piece of hardware which replaces the role of the traditional hub in which all nodes on a LAN are connected. Switches store MAC addresses of all nodes within an internal database, which it uses to perform its direct routing. Several manufacturers, including Cisco Systems, Linksys, and Netgear offer various types of switches with features such as 10/100-Base-T compatibility, gigabit Ethernet support, and support for Carrier Sensing Multiple Access and Collision Detection (CSMA/CD) which is ideal for high-traffic networks because it will queue connections and detect when packets collide in transit.
An emerging issue for enterprises today is that of mobility. Remote workers, field technicians, and executives require portable solutions, including laptops, Personal Digital Assistants (PDAs), and wireless access to network resources. The IEEE has established a standards body for the 802.11 wireless specification, which establishes standards for wireless data communication throughout all industries. The current standard in practice today is the 802.11b specification.
The 802.11b specification is actually a group of standards governing wireless communication and access control at the 2.4 GHz communication band. This specification has already been adopted at an industry level, and several vendors market 802.11b (also called Wi-Fi) access and compatibility as a value-added feature of their core offerings. Consumers have also embraced the standard for small-office/home-office (SOHO) networks. The popularity has also extended from LANs to MANs (Metropolitan Area Networks), especially in populated areas where a concentration of wireless access points (WAPs) are available. There are also wireless Internet service providers (WISPs) that cater to frequent travelers who require broadband Internet access to conduct business remotely.
The 802.11b specification allows for direct, peer-to-peer connections between nodes with wireless NICs. This loose grouping of nodes, called an ad hoc network, is ideal for quick connection sharing between two or more nodes, but introduces scalability issues that are not suitable for long-term wireless connectivity.
A more suitable solution for wireless access in fixed structures is to install one or more WAPs that connect to the traditional network and allowing wireless nodes to connect to through the WAP as if it were on the Ethernet-mediated network. The WAP effectively acts as a bridge router between the nodes connected to it and the rest of the network.
Although wireless networking is comparable in speed and certainly more convenient than traditional wired networking mediums, there are some limitations to the specification that warrants thorough consideration. The most important of these limitations is in its security implementation.
In the excitement of successfully deploying an 802.11x network, many administrators fail to exercise even the most basic security precautions. Since all 802.11b networking is done using high-band radio-frequency (RF) signals, the data transmitted is easily accessible to any user with a 802.11b NIC, a wireless network scanning tool such as NetStumbler or Wellenreiter, and common sniffing tools such as dsniff and snort. To prevent such aberrant usage of private wireless networks, the 802.11b standard uses the Wired Equivalency Privacy (WEP) protocol, which is an RC4-based 64- to 128-bit encrypted key shared between each node or between the AP and the node. This key will encrypt transmissions and decrypt incoming packets dynamically and transparently. Administrators often fail to employ this shared-key encryption scheme, however; either they forget to do so or choose not to do so because of performance degradation (especially over long distances). Enabling WEP on a wireless network can greatly reduce the possibility of data interception.
Relying on WEP, however, is still not a sound enough means of protection against determined malicious users. There are specialized utilities whose purpose is to crack the RC4 WEP encryption algorithm and exposes the shared key. AirSnort and WEP Crack are two such specialized applications. To protect against this, administrators should adhere to strict policies regarding usage of wireless methods to access sensitive information. Administrators may choose to augment the security of wireless by restricting connectivity to SSH or VPN connections, which introduces an additional encryption layer above the WEP encryption. Using this policy, a malicious user outside of the network that cracks the WEP encryption has to additionally crack the VPN or SSH encryption which, depending on the encryption method, can employ up to triple-strength 168- or 192-bit DES algorithm encryption (3DES) or proprietary algorithms of even greater strength. Administrators who apply these policies should certainly restrict plain text protocols such as TELNET or FTP, as passwords and data can be exposed using any of the aforementioned attacks.
For administrators who wish to run externally accessible services such as HTTP, email, FTP, and DNS, it is recommended that these publicly available services be physically and/or logically segmented from the internal network. Firewalls and hardening of hosts and applications are effective ways to deter casual intruders. However, determined crackers will find ways into the internal network if the services they have cracked reside on the same logical route as the rest of the network. The externally accessible services become what the security regards as a demilitarized zone (DMZ), a logical network segment where inbound traffic from the Internet would only be able to access those services in the DMZ. This is effective in that, even though a malicious user exploits a machine on the DMZ, the rest of the Internal network lies behind a firewall on a separated segment.
Most enterprises have a limited pool of publicly routable IP addresses from which they can host external services, so administrators utilize elaborate firewall rules to accept, forward, reject, and deny packet transmissions. Firewall policies implemented with iptables or dedicated hardware firewalls allow for complex routing and forwarding rules, which administrators can use to segment inbound traffic to specific services at specified addresses and ports, as well as allow only the LAN to access internal services, which can prevent IP spoofing exploits. For more information about implementing iptables, refer to Chapter 7.