Wow...what's with all the spam lately?

i already mentioned to set up a fake captcha image and write the actual captcha code into the normal info-text or wherever else =). i would be surpised if a bot is able to see though this trick. would be an easy solution… dunno if it really works but is might :smiley:

Same with swapping the “next” and “back” button XD

There are many things that may help, but its a question if the new forum members read the text above the captcha all the time. TBH: I don’t :stuck_out_tongue:

Regards, Bigfoot29

just make the in a way it’s not possible do oversee. bdw… you could use a image with next/back on it and write it the wrong way in the code with the right link-targets of course. if you are too stupid to see a captcha in a form like this…
IGNORE THE IMAGE- ENTER THIS CODE
A7Fb0
IGNORE THE IMAGE!!
then i wonder how they found the panda page at all^^
some kind of idiot-test :unamused:
if bots get thorugh… and people dont… well no comment…

greetings
thomas e

I’ve almost finished my CAPTCHA program, but the saved screenshots come out garbled. Here is the relevant code:

loadPrcFileData("", "window-type offscreen")
import direct.directbase.DirectStart

 . . .

def SavePic():
    base.screenshot('captcha.jpg',defaultFilename = False)
    sys.exit()
taskMgr.doMethodLater(1,SavePic,'Save Task',extraArgs = [])

Why doesn’t this work?

Hmm, I don’t know. It seems right. Garbled in what way?

David

It comes out looking like this:

I don’t understand. When I comment out

loadPrcFileData("", "window-type offscreen")

It pops up a window first and works fine.

Weird. I tried again today. Same code.

This looks like pieces from the Blender GUI. I had already closed Blender before taking this shot.

It almost looks like the screenshot() method is looking at the wrong memory address or something.

OK, there have been some issues with offscreen rendering in the Panda3D 1.2.x branch, but I think they have been resolved on the trunk. It may be that you’ve run into one of those issues.

What version of Panda are you using? If possible, you may do well to try to get the latest source and build it yourself.

If you are on Windows, you might first try switching between OpenGL, DX8, and DX9. One of these might work where the others don’t.

Another thing that might be worth a try is putting:

prefer-single-buffer 1

in your Config.prc. It looks to me like you’re grabbing the framebuffer from uninitialized memory, and maybe it’s getting the front buffer of a double-buffered frame. Telling Panda to use a single-buffered frame might solve this problem.

David

i just had a rough look at how these captcha recognition bots are working.
and after i found this on the web
http://forum.poland.com/index.php?s=1985db6d1294453f1c1c22b391bef573&showtopic=3146&pid=5783&st=0&#entry5783
it is highly possible that even very well made captcha images are read correctly.even if they are made from 3d objects.

instead i had a brilliantly stupid idea how to get one step ahead again.
:exclamation:the basic idea is to provide more information than actually needed.
since the actual recognition is good , a bot would easyly get all letters, but they simply too much.
to get only the needed letters you could use small arrows to point at them, or use colors and tell the user to only read a certain color. like my small example

the captcha would be “eMorts”

another idea from the captcha-craking side would be to use humans to decifer captcha-images. since it would be not economical to pay them they want to offer free por* to the users for reading captchas.
a more than fearsome idea which is doomed to work awfully good!
getting rid of them would be very difficult. using ASCII-art arrows made from / to point to the wanted letters in an image, or setting up a code-chart you receiver via emaili would be other solutions. but later ones aren’t very user-friendly anymore.

I like the colour idea… bots will possibly be able to sort that out soon, but hey, its an option :slight_smile:

Regards, Bigfoot29

I replaced

loadPrcFileData("","window-type offscreen")

with

loadPrcFileData("", "window-type offscreen\nprefer-single-buffer 1")

Still looks wrong. When I try

loadPrcFileData("", "load-display pandadx8")

(or dx9) with the offscreen it just crashes. “pandagl” was the default, but it’s garbled whether I define it explicitly or not.

I’ve been using Panda3D-1.2.3, which is from the latest Windows installer available for download from this site.

OK. Might be a problem with your graphics card, or a problem with Panda3D 1.2.3. What graphics card do you have, and are you using the latest driver?

If you’d like to post a link to your application, I can try it on my computer with the latest (CVS) version of Panda.

David

The script isn’t long (only 45 lines of code,) so I justposted it in Showcase

I’ve been using a GIGABYTE MAYA RADEON 9000 PRO graphics card. Driver version 8.231.0.0 (dated 2/21/2006). If there is a later version, I haven’t found it.

OK, yep, it was a bug in Panda. In fact, it was still present (!) in the CVS version of Panda, but I just checked in a fix. This fix will be included in a future release of Panda, or you can get the latest and build it yourself if you need it now.

You guys are getting good at finding Panda bugs lately!

David

I asked once and got no reply not even no thanks so Ill ask again…i’d be more than willing to peruse forums and snatch spam’s if someone is willing to trust me with admin access…I do peruse forums often so its likely I could be of help…no one should have to see these annoying spams if someone is willling to offer assistance and at least try to catch some of them until a sollution is agreed upon.

cheers
neighborlee ()

I was just thinking… maybe the whole idea of “capcha” isn’t the right solution. What if, instead of solving a visual puzzle, you just had to answer a simple question? Something like this:

Q: what color is the sky on a sunny day?

A: blue

The board admin would need to have a configuration page where he can type in the question and the answer. It would simply be a matter of trying out different questions and answers until you find one that most humans can solve, but most AIs can’t.

quite a nice idea… but bad luck if you never come out of you window-less room to see what color the sky has on a sunny day… but thats what wikipedia and google are for :wink:

Easier said than done. Most such simple questions would have more than one “right” answer, but the testing computer would only recognise the exact string explicitly stated in the configuration page.

English isn’t the native language of many of the forum users. The one “right” answer to a “simple” question proposed by an American board admin might be obvious to another American, but ambiguous to someone else. Therefore it might be as difficult to program and AI to recognise the possible “right” answers as it is to program an AI to guess them.

neighborlee: I think we can need some more referees. Lets see.
Edit: too few referees isn’t the problem if you can ban the spamming. We don’t have kiddies here that need a huge support base. But one more ref might be appropriate. As stated, lets see. :slight_smile:
Edit2: Ok, you got mail. :slight_smile:

in general:
The question-answer thing is also a good idea. But the problem is also mentioned already. The issue will be to find questions that can be answered without to even think about it. But even simple question could be answered wrong. How many fingers does a humans hand have?

Whats the desired answer? 5, because thats the overall ammount of “fingers” or 4 because a human hand has 4 fingers and a thumb?

Plus: That means that we need an php programmer to directly implement that into python (using a DB to gather the questions/answers from)

and a python project might lack php coders - I have seen huge portals written using python, so why use php. :wink:

Regards, Bigfoot29

I dont know about of PHP, but I might be able to find someone that does…and yup I got mail :wink:

cheers
neighborlee()