How I Hacked Snapchat’s Dumb Anti-Robot Security In Less Than 30 Minutes

I wake up this dawning and saw an article detailing Snapchat ’s new check system plan to aid trend back the snap spam that seems to be slowly infiltrating the service .

You may not have see it but if not I will sum up it for you . They fundamentally have you choose from amongst a lot of image , identify the ones that have the Snapchat ghost to bear witness you are a person . It is kind of like a less annoying CAPTCHA .

The problem with this is that the Snapchat spook is very particular . You could even call it a template . For those of you conversant with template matching , which is what Snapchat is enquire you to do to verify your human beings , it ’s one of the leisurely tasks in computer vision .

Hostinger Coupon Code 15% Off

This is an incredibly big room to avow someone is a person because it is such an loose problem for a computer to solve .

After I take this , I spent around 30 minutes write up some codification for make a computer do this . Now there are many ways of solving this problem , HoG credibly would have been well or even colour thresholding and PCA for blob realisation but it would take more time , and I ’m lazy ( read : efficient ) . I ended up using OpenCV and going with unproblematic thresholding , SURF keypoints and FLANN jibe with a uniqueness test to determine that multiple keypoints in the preparation epitome were n’t being singularly agree in the examination image .

First , I extract the different images from the slideway above , then I threshold them and the trace guide to find objects that are that colour . Next , I extract feature spot and descriptors from the test image and the templet using breaker and match them using FLANN . I only expend the “ best ” matches using a distance metric and then check all the matches for singularity to affirm one feature in the template is n’t fit most of the test feature article . If the uniqueness is high enough and enough features are found , we call it a ghost .

Burning Blade Tavern Epic Universe

With very little try , my code was able-bodied to “ find the ghost ” in the above exercise with 100 % accuracy . I ’m not saying it is perfect , far from it . I ’m just saying that if it takes someone less than an hour to train a computer to break an example of your human verification system , you are doing something wrong . There are a ton of way to do this using computing machine vision , all of them quick and effective . It ’s a numbers plot with computers and Snapchat ’s verification system of rules is suffer .

This mail originally appearedon the source ’s blog , and it was republished with license .

Steven Hicksonis an avidtechnical blogger and current graduate bookman investigator at Georgia Institute of Technology . He graduated magna cum laude with a level in Computer Engineering from Clemson University before moving on to the Department of Defense . After confer and work at the DoD , Steven decided to quest after his PhD with a focus in computer vision , robotics , and embedded systems . His open source libraries are used the world over and they have been featured in billet such as Linux User and Developer Magazine , raspberrypi.org , Hackaday , and Lifehacker . In his free time , Steven like to rock music climb , political program random bits of computer code , and bet Magic : The Gathering .

Ideapad3i