Or that image verification code thing?
I'm glad the report(s) are helpful. I was half under the impression you would just deal with them as you noticed and any report would just be an extra pm to read amongst many other jobs the admins would have to do. Like it might have been an inconvinience as you knew about these and were dealing with them as quick as possible.
I believe we already have an image verification (tell me if I am wrong, its been a while).
The problem is that the people who program bots are not stupid and have managed to "crack" the image algorithms used for many popular websites and web programs.
If we use any image verification, it is probably the standard vBulletin image verification system which has probably just been cracked.
*tangent time for those that are curious or bored*
I think they consider the image generation system "cracked" if the bots can guess the image code correctly 25% of the time, there have been a couple interesting ideas I have seen floated for how to make the algorithms harder
- There is a project going on to take actual books and convert them to text files on computers, sometimes their Image->Text algorithm fails (and recognizes a failure) and flags it, one group is working on figuring out which words / phrases it fails the most on and the theory is that if state of the art Image->Text converters fail on these words / phrases then an image reading bot should fail as well.
- An idea I had, I would imagine that other people have had it as well, is to implement how they do spam filtering but for the image generating algorithm. The way spam filter works is that it is a pattern recognition program that learns new patterns as more emails are reported as spam. Well since the majority of image failures are generated by bots failing at converting images, find the images it fails the most on and generate hybrid images from the ones the bots failed at, therefore presenting an always moving target for the bot makers to hit as the image generating program starts generating more complex images, of course you also have the problem where the images get to complex for even humans to read.... heh
As for the Question / Response plan, I think for our local site that could work out, we just need to come up with a list of questions / answers that should be widely known, or easily figured out. The bots follow scripted procedures so if we just switch up the process a bit then they should be stymied.
Like, "What is Seven times 3?" or "Is a circle round?", "Type the word heart in the answer box."
The real problem though is, which is less work, Having mods approve all new users or implementing a new registration system... I guess its up to the Site admins (DrPepper and DJive) to figure that out.