How to realize the class ExeVisualizationTests？ about secml_malware HOT 3 CLOSED

Hello!

Thank you for using SecML Malware!
So, there are two methods: plot_code_segment and plot_contribution_header_histogram.

The first one produces the heatmap with the single contributions, while the second produces the histogram.

I removed from this release the computations of integrated gradients, as it was difficult to generalise to other models that were not MalConv, but I could add that in the future.
To replicate it, you can use the code from the original repo of Integrated Grandients!
Remember that MalConv is differentiable up to the embedding layer, so you need to compute them only up to that.
Then, you sum up th contributions of each embedding, and then you have your vector of contribution to pass to the functions mentioned above!

from secml_malware.

Hello!

Thank you very much for your answer!
I know how to achieve it through your detailed description. I'll try it right away.The SecML Malware is a great tool for my follow-up research.I hope I can seriously learn the implementation of it, which helps me a lot！

Let me thank you again!

from secml_malware.

Thank you for using my library 👍

from secml_malware.