Bitcoin Forum

What if a faucet used webspeech API to protect against bots?

The faucet could have a box, a shape, font color whatever and it be a certain color, say green.
Then the user must click a button then speak the color of a specific object in this example 'green', in order to move to the next page.
A sortof speech captcha.

I think this would be relatively easy to set up, what do you think? I think it could really throw the bots for a loop, at least for a little while.


Article:  https://hacks.mozilla.org/2016/01/firefox-and-the-web-speech-api/?utm_content=buffera9386&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer

API Documentation: https://dvcs.w3.org/hg/speech-api/raw-file/tip/webspeechapi.html

Example: http://mdn.github.io/web-speech-api/speech-color-changer/

What's to stop someone from just scraping the web of audio files and putting them into a db to execute when said color, object, whatever is asked for?  I'm sure you could find many audio clips of someone saying the color green and just use a random one each time.

Also I would be curious to know what percentage of the people with internet access have a microphone.

It could change to different topics, like "a cow goes..." there are lots of different variations to protect against that. It would also take a lot of resources to compile a bot attack against that.

The access to a mic is a big concern that I didn't take into account....

What if it were reversed, and speech gave instructions like "click in the third box in the second column" and there was a grid of boxes for the user to click?

That's sort of similar to the audio version of Recaptcha, where it asks you to type the numbers you hear people speaking.  And just like recaptcha in order for it to work you would need a large range of audio so people can't match a hash to an audio clip and correspond it to a specific action.  I recall even Recaptcha had issues with people using speech recogintion software to decipher the numbers as it was easier to solve than trying to ocr the visual images.  This is all assuming someone doesn't just bypass the whole audio process and figure out what the correct box choice is programatically.

Not sure what you mean there... no audio clips would be needed, the webspeech API allows you to insert any text you decide server-side to be spoken outloud on the client side, it's synthesized speech. It would be no different than storing text in a db.  
If the user selects the wrong box than a new puzzle starts, so they couldn't just start clicking random boxes or expect it to be the same box each time.

OK, Satoshi! Give me some of that digital staff of yours! Please.  ;)

The idea is nice but have mix feelings if it would be able to solve the problem of bots without interfering human users in terms of making the whole problem a complex one. In terms of short term effect I think it could succeed but in the long run, programers will surely make a way to bypass that as well, no matter how wide their range gets. Also as you said, the use of mic, could complicate the situation since some don't have any and have to provide just so they could earn fractions of BTCitcoin. Perhaps if there was a way that could be implemented in a way that each address user have to validate it's identity so only one user is tide to each address and then bots could be eliminated although it requires separate data base and it will get anonymity from those who want to remain one.

I don't think its good idea. First it will ask visitor to ON microphone plugin. It will also have language problem most of faucets visitors are non-english so you will have to include major languages. Also it can easily be programmed by bots. Color recognition by browser is so simple method to code.

This is not a crazy idea . I don't have micro phone so I'll not be able to solve that captcha ?
many users like me don't have micro phone .

Please choose your language asks the captcha

Hundreds of them, that is a big server database to have.