On 9/04/13 03:48 AM, [email protected] wrote:
Just saying...

They have signatures now, but there's no way to effectively audit them or 
expire them.


The question is, why? If you can answer that effectively, then you might be right.

Let me put these stumbling blocks in your way.

Firstly, signatures, if they are to be last over time, have to protect or mean something. In an ephemeral protocol, they protect a secret until it is thrown away, seconds or hours, no more explanation needed. In the other extreme, a long term digital signing protocol, they protect a really big public statement over years. So, your idea might have merit. But even there, a successful signing protocol always leans on immediate escrow of the signature and keys. Aka timestamping. No expiry needed, then, but there is a lot of load put on revocation.

Let's look more closely at the application: SSH does ephemeral key protection, not anything more serious. It is used by sysadms on their own machines, generally. It's their value at risk, and they are in control. They can audit keys by looking at them, both ends, and the entire cycle is under their control. They can expire them by generating new keys, and afaik, most sysadms do roll over their keys from time to time. It's really easy, one command to create. No money, no time required, except the scripting to promulgate.

If you look at say where they do use audit & expiry, the CA business, you'll find other circumstances: The users' value is entirely at risk, they are liable for everything, for which privilege they pay a fee. Yet the CAs are entirely in control, with no value at risk, for which they charge a fee. The jurisdictions are multiple: users, vendors, CAs, server vendors, server users, and committees, all with particular interests. Money and time are significant. Even the signatures mean different things: code-signing != SSL signing != client-login != client-signing.

One thing should be clear: just because the CA business does X, doesn't mean that anyone else should.



iang
_______________________________________________
cryptography mailing list
[email protected]
http://lists.randombit.net/mailman/listinfo/cryptography

Reply via email to