DEV Community

😛 Useless Code for Tinder 😛

0xkoji on July 24, 2019

As some of you know, sometimes really need to write code for completely useless things 😛 1. Open Google Chrome and Developer Tools ...

Read full post

Vaibhav Namburi • Jul 24 '19

Take it up a couple notches, extract the image, run image analysis on common features you like. A nice simple ML bot and train it to your needs.

Then let it run in the wild. We did this for a friend, funny AF

Daniel Waller (he/him) • Jul 24 '19

I don't know man...I get that its a fun project for shits and giggles.

But mining pics from dating sites (albeit publicly available) and using them for feature extraction/image analysis without their owner's consent feels kinda wrong

Vaibhav Namburi • Jul 24 '19

That's and absolutely fair point.

To note this was done in the frontend using tensforflow, no data store etc.

But nonetheless good point, and something to reflect on, definitely not right thing to do at all

Médéric Burlet • Jul 24 '19 • Edited

When I was bored I made a tinder finder.

Tinder gives you distance between you and target
You can change your position via parameters in the web request
With a little algorithm and automatically changing position after few location jumps you can pinpoint the target's location

Hint: Used a variance of a trilateration algorithm

Ouput:

0xkoji • Jul 24 '19

nice!

Rehan van der Merwe • Jul 24 '19

This is scary can you really change your position in the params of the web request? Also how accurate was this?

Médéric Burlet • Jul 24 '19 • Edited

Yep just open the dev tools on tinder web and then look at the request. You can also use location spoofer plugins for your browser.

For accuracy I guess it would go down to 1km as this is the shortest distance tinder shows. If it was more accurate Tinder would face some issues.

However without knowing how tinder handles distance we can't know how precise it is.

It still is a fun project to code however

petercour • Jul 24 '19 • Edited

Nice! The js is cross platform.

Instead of random, what I did was use adb to fetch the profile image, then run it trough face detection algorithm and several others. The machine learning algorithm would then decide like or not like based on features.

It wasn't perfect, as the face detection wouldn't work if faces were turned. But otherwise "saved me time". It was all automatic, so it would do that for each swipe.

Then I went on some dates, that the AI had effectively decided for me