Foolbox FGSM Adversarial Attack
Upload an image and apply an adversarial FGSM attack with your chosen epsilon.
Original
🧪 Lab Instructions: Foolbox FGSM
- Upload an image you'd like to attack (PNG or JPG).
- Set the
epsilonvalue to control perturbation strength. - Click "Run FGSM Attack" to generate an adversarial version of the image.
- Compare side-by-side output, and click "Download" to save the adversarial image.
🔍 What is FGSM?
The Fast Gradient Sign Method (FGSM) is a simple and efficient adversarial attack. It works by adding small perturbations to an image in the direction of the gradient of the loss function with respect to the input. This causes the model to misclassify the image while making only minimal changes to it.