Skip to content
This repository has been archived by the owner on Jun 15, 2023. It is now read-only.

Latest commit

 

History

History
19 lines (11 loc) · 1.74 KB

README.md

File metadata and controls

19 lines (11 loc) · 1.74 KB

Flying Pig I

Please proceed to the flying_pig folder.

You will modify the pig.jpg image so that it will be misclassified into a plane by a model. (We already provide a pre-trained model to classify images into corresponding classes.)

Your first job is to create an adversarial image by interpolating two images. You can find pig.jpg and plane.jpg. You can try interpolation of the two images in various ways, e.g., computing the mean of the two images, scaling the plane's pixel values by a factor of two and subtracting the pig data from there, etc.). We provide template.py and at the first TODO, you can interpolate the two data (pig_data and plane_data). Running the template.py will create flying_pig.jpg in the same folder; then, you can run launcher and choose option 1 to get the flag.

Good luck.

Flying Pig II

Please be in the flying_pig folder.

You will modify the pig.jpg image so that it will be misclassified into a plane by a model. (We already provide a pre-trained model to classify images into corresponding classes.)

Your second job is to create an adversarial image by running an adversarial-example crafting algorithm. We offered the algorithm. It will construct perturbations to the pig image iteratively, such that the resulting perturbations, when added to the original pig image, make the pre-trained model misclassify it into a plane class. We provide template.py and at the first TODO, you need to put the class numbers to the location where marked as [0] (e.g., from zero to [143]). You can choose which of the 341 and 404 should be placed in the zero location.

Running the template.py will create flying_pig.pt in the same folder; then, you can run launcher and choose option 2 to get the flag.

Good luck.