Hi Jow Forums, /qa/ here

Hi Jow Forums, /qa/ here.
How viable would ot be to make and implement a game like, say, a touhou game as a captcha and how effective would it be against spambots?

Attached: 1565298093260.png (535x610, 328K)

Other urls found in this thread:

github.com/bakape/captchouli
twitter.com/SFWRedditGifs

>game
wrong board buddy, manchildren belong on

It wouldn't be very effective because the patterns of the bullets are always the same
Also go back to /v/

>How viable would ot be to make and implement a game like, say, a touhou game as a captcha
fun/10

>how effective would it be against spambots?
Very ineffective. Captcha works because you have to do something that humans are good at but computers are bad at.
Touhou seems like something computers can be easily good at, it requires mostly reaction time and repeating patterns which computers are really good at.

This. It would be a reverse captcha. Only bots get to post...
Could be fun

oh i remember this thread

Implementing it wouldn't be too difficult, you should be able to hack something playable together in a couple of hours. It wouldn't be effective at all though since playing the game would be completely optional since a bot could just send back whatever win response you have. That is unless the game was running on the server-side and you just sent over keystrokes in which case I imagine it'd be unplayable and near impossible to beat.

Not very effective. Use this instead.
github.com/bakape/captchouli

Attached: sample.png (463x525, 378K)

lat is based
might as well change the board software to meguca too

Attached: 1542765513167.webm (486x528, 2.31M)

>How viable
client-side_injection.js/10

Literally dont know half of these, and thats after spending years and years on /a/
Shit solution, even for its intended purpose

Useless, today very basic Deep learning can outperformance this kind games.

Better will be user select between IA generate post and real post 4channel.

kys you autistic faggot.

You’re an absolute newfag, you shouldn’t be posting.

wtf is with the ADHD mouse movement

Attached: tards.png (398x97, 113K)

I think this would be way more effective than the googlel garbage

the problem I see with using danbooru's image bank is that boorus have a finite number of images and that can be used by anyone, including someone who wants to break the captcha by developing a bot that can recognize the origin of the images that appear in the captcha.
personally I think that using software like thiswaifudoesnotexist.net or make.girls.moe that generates anime girls with neural networks would be a better alternative, because there would be an infinite number of images that can't be backtraced, not to mention that distortions in the images created by errors in the NN could serve as an additional filter to confuse bots that try to recognize the faces.

Should be required to post on Jow Forums and the only approved captcha system.

That wouldn't work because then how does the system recognize which girls are valid without handpicking them in the first place?
Let's say the system asks "pick all the girls with blue eyes", that then means the system is able to recognize girls with blue eyes which also means any bot could recognize girls with blue eyes and make the association and bypass the captcha.

Unless I'm not thinking of all the possible variables here...

in the case of make.girls.moe you can choose the characteristics of the girls that will be generated (hair color and eyes, hair length, accessories, etc.). that could be used to identify them.

As much as I love this system, the fact that this webbum starts with a captcha with no correct answers bothers me.