This time we introduce a new type of gradient based attack, in order to brute force an image classification app \(can be shaped and used for any input of course\), the BIM, or Basic Iteration Method.
It’s recommended to see at least the explanation in the [**introduction challenge colab Notebook**](https://colab.research.google.com/drive/1lDh0oZ3TR-z87WjogdegZCdtsUuDADcR)\*\*\*\*
To go deeper on the BIM topic:[ https://arxiv.org/pdf/1607.02533.pdf](https://arxiv.org/pdf/1607.02533.pdf)
As usual we will provide only the A.I. attack core part, it’s up to you to complete the tool and blending it with PT techniques, depending on the situations.
{% hint style="info" %}
Remember, in those kind of scenarios, in order to mime real-based attack applications, we don’t have the exact model to fool or the image target in which we would like to transform our image. That’s why, in order to overcome this issue, we must blend our core script, with a bruteforcer logic, accordingly to the application responses we want to fool.
