Two years ago, the bulk of the questions I got from enterprise clients were about security. No surprise there - if radio signals propagate throughout the intended coverage area (and frequently beyond), what's to prevent unintended receivers from grabbing stuff they shouldn't have? And what's to prevent unauthorised network access, the other key element of network security?

Wireless was looking like an open door, with "steal me" or "come right in" written on it in big shiny letters. The fundamental flaws in Wired Equivalent Privacy (WEP), the first 802.11 security standard, didn't help. I'm surprised that WEP hacking never became an Olympic sport, what with all the free tools available on the Internet. Crack a key in 15 minutes or less or your money back.

WEP was quickly replaced by Wireless Protected Access, which is basically still WEP but with the keys changed on a rapid and regular basis. We recommend WPA, which is fully supported in Windows, as the base-level wireless security all enterprises should adopt. If you want more, the recently approved 802.11i standard offers an entirely new encryption algorithm based on the Advanced Encryption Standard (AES), and this one is really, really tough to break.

But there's a fundamental flaw, again, in using just wireless security capabilities in an enterprise environment: It only affects the wireless part (the airlink) of the value chain. Because there's really no such thing as wireless security, we must think about overall, end-to-end network security. If we solve that problem, as it turns out, we can almost ignore the wireless security elements once basic (WPA) precautions are in place.

So, what we recommend to enterprise network managers deploying wireless LANs is the following:

-- Have a security policy. First and foremost, have a policy regarding all sensitive data. It should state who has access and under what circumstances, and what procedures need to be in place. From this, an appropriate technology solution can be considered, planned and implemented.

-- Encrypt and authenticate at the server. All sensitive data stored on the enterprise network needs to be encrypted to prevent theft particularly in the case of physical compromise. Similarly, one needs to authenticate before access to the data is granted. I'm a big fan of two-factor authentication; hardware tokens will become much more cost-effective over time, and I expect they will become the norm.

-- Encrypt and authenticate at the client level. Similarly, and I know this is a pain, any mobile device holding sensitive data must also encrypt that data, and users must authenticate with the device. Entering a PIN is a pain on cell phones, but it's necessary in many cases.

-- Use VPNs everywhere. Finally, use a virtual private network to ensure that data is secure end to end. I think Secure Sockets Layer VPNs will dominate as Web services become the norm, but there are many good choices available.

Using the above philosophy, data can appear in the clear only to authorised users, and the theft of information and unauthorised access become very difficult indeed. Of course, there is no such thing as absolute security, but the above works pretty well.

Note also that during the WEP era, our primary concern was the casual hacker, who was trying to get into networks for reasons of mischief, bragging rights, intellectual curiosity or a pathological need for Internet access. The real threat is the professional information thief, who gets paid to commit crimes and isn't likely to be found within range of an access point with a Pringles-can antenna, waiting for the top-secret strategic plan to come flying by.

Rather, they're going to look for weaknesses elsewhere in your network value chain. That's another reason I like the above approach - it's end to end. And as a bonus, it will work with any network, including wide-area wireless services.

As it turns out, the constant work-in-progress nature of wireless - um, I mean network - security has historically been one of the two key challenges to the adoption of wireless in the enterprise. I hope you'll agree with me that this problem is at least manageable with the end-to-end approach. Next week, we'll look at the other problem, the one that has yet to, and may never, be solved.