Craig Burton has penned another crystalline piece called How to Spot an Unnecessary Identity Fail (after his previous piece, How to Divine the Bovine, this is starting to sound like a field guide to identisaurus). His key point: we’ve had asymmetric key cryptography for 30+ years and we are still storing usernames and passwords on servers where they can be ripped off.
What’s wrong with this picture?
In an IM session with him I pointed out that while moving entirely to asymmetric keys is a giant security win (because your private key is never stored on a server, at least not unencrypted), the problem has always been the usability of foisting private key management on the user (which is the only place it can truly be and still retain the full security advantages). Even Microsoft with their design for Information Cards (which are the closest we’ve ever come to full asymmetric key-based security infrastructure) never fully solved that problem.
Craig’s point is: that’s where the innovation needs to happen. Focus on that one fracture point and you can split the entire Internet security boulder.
And if you use password digests, or fancy split-key recovery protocols such as those Ben Laurie has been working on at Google, you end out not storing a secret anywhere except in a user’s head.