A more adaptable approach might be to rewrite this as a client-server application, though there is the risk of creeping-featurism if this is not handled carefully.

Essentially, if you can store a "secret key" on both the client and the server in a pseudo-secure manner (i.e. both machines are trusted, or trusted enough), you can use a simple MD5-style hash to authenticate with a challenge-response system. The MD5 "secret key" is really just some stuff you made up, like a whole pile of random letters.

The "server" part is a CGI program that asks a "question", presenting a series of random characters that are to be encrypted by the remote client. The "client" is the program that uses LWP to access the server. It downloads the question ($response->content()), encrypts it with the key, and sends it back for verification. If the crypt() passes the muster, the server records the address that the request originated from, plus any other data you might have sent with it (i.e. your "real" address, should it be different, or what have you).

If you're feeling extra lazy, you can use LWP::Simple instead of going whole hog. Also, forget sending HTML back and forth, just use HTTP to mule your data raw.

Client/server package is probably only about 20 lines of code, when you get right down to it.