Benjamin Franklin was undoubtedly thinking of 802.11b when he famously said those who would sacrifice freedom for security deserve neither. The convenience and low price of wireless networks has lead to their uncritical adoption, despite the fact that everyone knows they are fundamentally insecure, and while wily hackers with Pringles antennae are more often to be found in the media than outside your offices it remains true that wireless networks are unacceptable in secure environments.
A range of updates are needed to the 802.11 set of standards to make wireless secure, and the IEEE has bundled them together in 802.11i, currently nearing completion. It adds two main blocks of improvements, improved security for data in transit, and better control of who can use a network.
It's often overlooked that security on wired networks is also something of a novelty, people assume because a LAN is entirely contained within an office it is therefore safe. That's dangerously wrong, on-wire security being every bit as important as wireless. The basic building block of Ethernet security is 802.1x, approved in June 2001, and 802.11i is designed to integrate this with wireless.
Three components of security
Security has three main components, authentication to confirm both sides of a transaction are authorised to exchange data, encryption to protect the data in transit, and key management to allow both sides to read each other's messages. Of the three, encryption gets the most press because it's easy to comprehend -- it's also the easiest to get right. 802.1x doesn't define any encryption standards, instead it handles authentication and key management. It can be used with any cipher, and with many authentication methods. At the heart of 802.1x is the Extensible Authentication Protocol, EAP. This describes how two network nodes can pass messages to each other asking for authentication -- the standard was first coined for dial-up authentication over the Point to Point Protocol (PPP). 802.1x adds EAP over LAN (EAPOL). When a network access server -- typically a router or a wireless access point -- detects a new client, it sends an EAPOL message requesting its ID. The client returns the ID, which the access server then passes to an authentication server -- commonly a RADIUS server. This then has a conversation with the client, the access point relaying the messages, until either the client is accepted and authenticated or rejected altogether. Until this point, the only access the client has to the network is as a generator and consumer of EAPOL messages -- nothing else is allowed. Roaming between hotspots
Once the client is recognised and accepted, the authentication server can also provide authorisation for different levels of access, depending on the client's ID. This opens up the range of services that the client can access from the port provided by the access server, as well as potentially setting quality of service, rate caps and other user limits. Note that the RADIUS server can be far away from the access server, perhaps even on a different network, which opens up the possibility of roaming between different service providers of 802.11 hotspots. 802.1x also specifies how keys are passed back to the client to be used in further network traffic -- how these keys are used is not specified, but how they are transferred securely is. It also sets a per-packet authentication key that can't be faked by a third party, maintaining authentication if the client roams to another port and preventing interception and taking over of a session by an intruder. In installations where there's no authentication server -- probably the case in most homes and many small businesses -- 802.1x can be used in pre-shared key mode, where every node has its keys set up explicitly by hand. The rest of the 802.1x features work, with the proviso that if the shared keys are ever compromised the security of the network is lost. When 802.1x was started, wireless networking wasn't nearly as prominent as it is now, so the standard is designed for a variety of wired networks only -- hence the need to incorporate it into 802.11i before it can be used for radio. 802.11i's data security additions include various encryption processes such as TKIP (Temporal Key Integrity Protocol) and CCMP (Counter with Cipher Block Chaining Message Authentication Code Protocol). TKIP's main attraction is a frequent update of the encryption key, TKIP can be added to an existing 802.11 interface by upgrading its software, while a version of the TKIP mechanism called SSN (Safe Secure Networks), has already been adopted by the WiFi industry group prior to the approval of 802.11i. This is a temporary measure due to the need for a fast fix to the broken WEP standard. CCMP is designed for future wireless LANs, as it needs more processing power than most adaptors and access points currently have to spare -- it uses a version of the Advanced Encryption Standard (AES), the current US Government approved method of encrypting data in transit. 802.11i remains under development
With most of the technicalities decided but some areas -- such as fast roaming between access points -- still receiving attention. Such is the pressure for secure wireless LANs that some systems are already available with pre-approval versions of the standard. These are better than non 802.11i systems, but deploying them without a guaranteed upgrade path to the finished standard has interoperability and security risks of its own. By the end of the year, the standard should be finalised and equipment available: then the IEEE will have done its part, and it will be down to system deployers and managers to configure and maintain adequate security. 802.11i is just the kit of parts to do the job -- network security only works when people do it right. For a weekly round-up of the enterprise IT news, sign up for the Enterprise newsletter. Tell us what you think in the Enterprise Mailroom.