TensorFlow Tutorial - Adversarial Examples

 

Notes by Magnus Erik Hvass Pedersen: https://github.com/Hvass-Labs/TensorFlow-Tutorials/blob/master/11_Advers...

In the previous tutorials, we have used various kinds of Deep Neural Networks for classifying images with varying success. In this tutorial we will see a simple method for finding so-called Adversarial Examples that cause a state-of-the-art neural network to mis-classify any input image to whatever class we choose. This is done simply by adding a small amount of 'specialized' noise to the input image. The changes are imperceptible to humans, but it fools the neural network.

This builds on the previous tutorials. You should be familiar with neural networks in general (e.g. Tutorial #01 and #02), and knowledge of the Inception model is also helpful (Tutorial #07).

Resource Type: