To hell with proprietary encryption algorithms

I sat in the front seat of a Mustang convertible, next to the driver. In the back seat sat The Third Man, who was demonstrating how easy it is to break into a wireless network using a laptop, Global Positioning System, wireless LAN card and free downloadable software.

We drove around Las Vegas the day before DefCon and found an endless supply of wireless networks. How do you break in? Reboot your computer, the wireless access point sees you, Dynamic Host Configuration Protocol assigns you an IP number, and you're a remote wireless node on the net.

In only two cases did we find networks that use the Wired Equivalent Privacy (WEP) algorithm. WEP is fundamentally useless because the 26-bit algorithm can be routinely cracked in less than four hours, again using downloadable tools. Why anyone would use wireless nets is beyond me, especially knowing that a break and enter is as simple as firing up Windows from a car or the nearest McDonald's.

So I talked to folks who live and breathe breaking through security and encryption. RSA Security just announced a $200,000 prize for the crypto-geek who can successfully factor impossibly large numbers. That reminded me that the older and weaker RC-4 algorithm was cracked by a distributed processing assault.

During the last decade, companies have routinely tried to crack proprietary cryptography. Visit www.elcomsoft.com for a host of products whose sole goal is to crack password protection on Microsoft and other major products - ostensibly to recover lost corporate files. Search for "password crackers" and you'll find every kind imaginable. The cryptography in Lotus Notes is another victim of aggressive and successful crypto-hacking.And things only get worse from there.

It turns out that major mission-critical, enterprisewide software packages are just as vulnerable to crypto-""hacks. Imagine if you found that your entire database was not really protected by "strong proprietary encryption algorithms," as the vendor claimed; or that your payroll system's password security was similarly vulnerable because the vendor figured it could out-design the best cryptographers in the world.

Later this year, a group of security professionals plans to release a study naming some top enterprise applications with screamingly weak cryptographic implementations. They are especially focusing on embedded cryptographic security for database applications.

This study will provide enough evidence of how weak these "strong proprietary cryptographic algorithms" are. But the authors will stop one step short of releasing the step-by-step methodology on how to crack them. The goal is to get vendors to 'fess up to their crypto-errors and then repair the hundreds of thousands of vulnerable systems deployed worldwide. No matter; soon enough hacks such as these become public knowledge, to the benefit of malicious insiders and external attackers.

I don't get it. As an industry, we have some pretty good cryptography out there. Whence comes the arrogance that applications vendors can do a better job than the best mathematicians and trained cryptographers the National Security Agency, Government Communications Headquarters and academia can muster?

We have the Data Encryption Standard (DES), which still provides a free and reasonably good, well-tested means of protection. Triple-DES, which is good enough for the banking community, is also free and thoroughly understood. The new Advanced Encryption Standard will take us a "guesstimated" 20 years forward, and RSA and Network Associates have stables of proven cryptographic methods. What's with this proprietary stuff?

Developing the best cryptographic algorithm is a battle the commercial software vendor should not enter. With the incredibly complicated mathematics, expensive and advanced technologies, and limited set of skilled humans, vendors are best suited to implement well-known, open source, tested and accepted cryptographic approaches.

My advice to user companies is to use approved and well-known public algorithms, not proprietary ones. Implement cryptography carefully, making key management your focus. Use trusted third parties for testing and evaluation.

If your vendor uses a proprietary algorithm for anything you want to protect, such as data and passwords, run. If your vendor claims that the proprietary algorithm is secret, run. If your vendor won't show you or the cryptographic community the engine that makes its cryptography so great, run.

Vendors: Stick to your strong suit, your business application. Hire proper crypto-geeks and let them do the job right. The alternative, as we may see in the coming months, may be terribly painful.

Schwartau is president of Interpact, a security awareness consulting firm, and author of many books, including Cybershock and Computer and Internet Ethics. Reach him at winns@gte.net.