You get a (mostly-reliable) IP address from a network (TCP) connection. If the client is connecting to the proxy and the proxy is connecting to your site, you'll see the IP address of the proxy.
As noted by others, most proxies will supply the IP address of the client in an HTTP header, but this isn't what will get logged to your web server logs (I think - maybe apache and IIS have that as an option?)
But for your case - I don't think you actually care whether someone being abusive is connecting via a proxy or not. If an IP address is being abusive, you probably want to suspend that IP address or take some other sanctions (disable signups from that IP, rate-limiting etc).
In fact, if the IP you suspend is a proxy you're hurting everyone else who might be legitimately using it, you should probably be slightly less willing to ban the IP if you happen to know it is a proxy.
If you're concerned about mass sign-ups, you may want to try other countermeasures. If you notice a spike in signups, perhaps disable them.
Some character recognitiion (CAPTCHA-style) appears to be the state-of-the-art test for a human at the moment, but I'm told (fairly unreliably, but the idea seems sound) even these are being foiled by people who (quite cleverly) re-present the captcha to humans visiting their own site. e.g. someone runs a porn site and to get some freebies people have to type in the answer to a captcha. So you can serve the captcha from your sign-in page to a human, get the answer posted back to their server and then on to yours.
Adding in a challenge/response over email is probably worthwhile too, if you're able to ask your users for an email address. |