Image Scaling Attacks are CRAZY!!! Hiding images in plain sight (Machine Learning)

With an Image Scaling Attacks an adversary hides a smaller image within a larger image. The small image becomes magically visible when the computer resizes the large image. This is possible when vulnerable code uses insecure interpolation. In this video we walk through the basics to help understand a set of scenarios where scaling might be vulnerable (especially in the machine learning pre-processing stages) and we also highlight how to mitigate the problem.
Blog: embracethered.com

[@MichaelRygaard]

OHH come on link to the code we we can try it

[@embracethered]

Oh sorry, I thought it’s in the video. The original research and tooling can be found here: github.com/EQuiw/2019-scalingattack