In recent days, Dark Reading's message boards have been buzzing with a debate on passwords, people, and the problem of security. This seems to be a perpetual or at least recurring discussion, and I have yet to put in my two cents (or 100 riel, in my case here in Cambodia). So I figure it's about time.
Single sign-on is the way this industry needs to move. Now, Im not saying that passwords are the way to go, and I realize that there are interoperability issues. Really, what we are talking about here is narrowing things down to three or four sets of credentials that are hard to forge.
We all recognize that Amazon.com is not going to get together with every major corporation and agree on using a common user database. It's just not going to happen. But most users I know would be happy to have three or four common passwords, instead of 10 or more.
It is entirely possible that somebody out there will provide a Web service that could provide reasonable authentication services for major sites. In fact, such services already exist. And there are several different options for single sign-on within an enterprise.
But here's the rub: Time has shown that single sign-on software is not trustworthy enough to become the one basket that most people want to put all their eggs in. Take the recent problems with MIT's Kerberos as an example.
Kerberos has been employed in large environments for years and years, is open source, and has been extensively reviewed by hundreds (if not thousands) of security-minded developers. And yet, years after the release of the software, we find that the RPC library has a serious bug that allows unauthenticated users to run arbitrary code on the Kerberos server. Oops. That cant be good.
On the other hand, password management tools have problems as well. Too often, they use poorly-implemented or proprietary crypto to manage the user passwords, and they are still vulnerable to hijack by compromise of the host machine.
Most password management tools are not portable, and they have a pretty awful failure mode. If I have 100 different passwords for 100 different systems, and my password manager fails, I am required to change passwords on all of those systems. Not an easy task.
So whats the solution? I dont know. Perhaps a portable (read USB-based), standards based (AES anyone?), multi-platform (Windows, Mac, and Linux at minimum) password management solution that includes a reliable online backup with a trusted third party. Oh, and of course it needs to be inexpensive and invisible (or nearly so) to the end user.
Any vendors out there willing to take the challenge?
Nathan Spande has implemented security in medical systems during the dotcom boom and bust, and suffered through federal government security implementations. Special to Dark Reading