Brainstorm.

Public-key cryptosystems are based on the idea of secret numbers and calculations that are "easy" to perform and "hard" to undo. E.g., a computer can multiply two big numbers together in a negligible amount of time, but it takes much longer, given that result, to find the two numbers that went into it.

I'd like to set up a lightweight security system for various low-security tasks that I want to do with a computer, but I want to be able to do the "easy" part without the aid of any special software. I want to be able to do it in my head or on paper without straining myself, so that I can use public computers to interact with my server. For example, I want to be be to post to this column just by sending an email to the server—from anywhere—but I want to secure it by adding some string of characters to the mail that no one else can forge.

Working together, can we think of a numerical computation or symbolic monkeying that's easy enough to do unaided by a calculator but is nonetheless modestly difficult for a stranger to undo, even with aid of a computer?

Let me sketch out how this might work. Let's say I have a secret, like a password. Maybe it's a 20-character string. I sit down with a piece of paper and write down the string, and underneath it I write the date and time. Then I start to do some recombinations and obfuscations and I end up with a string of a hundred characters—the signature. By including these characters in an email, my server can validate that I actually went through this process, starting with my secret password, and therefore the email is valid. The technique should also have the property that no one observing the emails should be able to work backwards to find my password, short of spending a modest amount of computer time—let's say two weeks on a top-shelf PC. By changing my password once a week I could stay ahead of any crackers' attempts.

Possible?

UPDATE: I think I've got something. More when I get back from stormy Rochester, NY.