Update problems

Asirra was mentioned in some emails that went around. I don't think Asirra is fundamentally any better than reCaptcha for the main problem. I suspect there's a high probability that rather than pure computer-based bots, the reCaptcha system was being broken by a Mechanical Turk style system that farmed the captcha breaking task out to humans. From what I understand that is common these days. Asirra is equally vulnerable to that technique. Assuming that is the case, the reason my modifications help, would be that rather than preventing the captcha-breaking itself, they make it incompatible with some automation systems used to farm out captcha breaking. I have nothing against using Asirra if people want to switch it to that instead of reCaptcha, but at least in theory I doubt Asirra vs reCaptcha makes much of a difference.

I'd prefer math+Asirra over just Asirra, because I'm pretty sure simply making the system on this wiki slightly unique, can break a significant subset of ageneric wiki-spamming setups. It's not like we're in the situation of wikipedia or other large sites where people would modify their spam tools specifically for this site.

I'm also pro-Asirra, but also content to leave our current setup if it's working.

I don't know if it's machines or people cracking the captchas. I know there exist people cracking captchas, but I'm not sure if we're a high profile enough target to make that worth while. I use Asirra on the BerryBots wiki and it's completely shut down any spammy registrations. (I think I was getting ~1 a day before that.)

From what I hear, the people-cracking of captchas are usually done by embedding the captcha from another site into a site for warez or porn, then the customers solve them to 'prove they are human' or whatever without realising they are facilitating spam.

I'd also rather do something custom, simply because it means that automated tools will never be able to crack it if they haven't been designed for it. It's only a matter of time before Asirra is cracked and we have revolutionary advances in object recognition =)

The cool thing about technological arms races is that from a participant's perspective, all progress is relative and temporary, but from an outsider's perspective, they can actually advance all of humanity.

Are you? I didn't know you were using it on the BerryBots Wiki.

I would still like to see if it by itself can completely mitigate the spam here on the wiki by itself (if only because I am curious now).

If you want, I would be okay with trying a switch to just Asirra for account registration, so long as we keep the UrlStopper extension to block external links from brand-new users. That way no damage can be caused. I could also set up UrlStopper to log events when it is triggered, to confirm whether random user names that registered try to spam. I'm feeling like behavior-based filtering like UrlStopper is really much more effective than "figure out if you're a human" systems can ever be anyway.

I agree. I would never recommend removing that safeguard. Since it has the added benefit of stopping human spammers.

Well, impatient human spammers anyway.

Is there another kind?