Exploring Adversarial Machine Learning

[adityaraj.arm]

Hi i am having issues in solving the evasion_assessment and poisoning_assessment.
In evasion assessment I have successfully evaded the model and got the model to label cock as laptop, but autograder labels still labels it as cock. the hint says add more evasion, so I tried changing the eta and n_masks but still no good.

In poisoning assessment, I successfully build the model but the autograder gives me output as ‘Bad model’. I even checked with the model if there is any problem but it is showing cat image in frog label.

Rest of the assessments I was able to do easily but on these two I have spent around 8 lab hours, and my lab hours are almost finished. So, if I can get an extension in lab hours that would be really helpful and some guidance on the problem I am facing.

thanks.

[TomNVIDIA]

Hello,

I have granted you 8 more hours. In the future, you can avoid burning the usage limit by pressing the “Stop” button when you’ve finished with a module. If you need to save your progress, you can download the necessary files from the lab.

If you have other issues with the lab, you will need to reach out to the DLI team directly. [email protected]