CryptoKnight

CryptoKnight copied to clipboard

While trying to run CryptoKnight, I'm able to draw the distribution and get the model trained. However, when I try to use knight.py to predict an executable, I get this:...

vsarunhah

In [the paper](https://www.mdpi.com/2078-2489/9/9/231), a distinct subset of cryptographic algorithms & opcodes were correlated which resulted in our model achieving 96% accuracy. It would be advantageous to explore different settings which...

gregdhill

The new model, previously introduced by [Nal Kalchbrenner](https://arxiv.org/abs/1404.2188), has been implemented in PyTorch, but due to time constraints for the original publication I implemented the `k-max pooling` and `folding` operations...

gregdhill

To enhance the portability of this framework, and due to the extensive number of dependencies, it would be good to provide a (pre-trained) [docker](https://docs.docker.com/get-started/) container. This could also be used...

gregdhill

The framework currently traces binaries linearly which is an inherently slow process. Compilation is relatively fast, but tracing each binary with Intel's PIN framework is by far the largest overhead....

gregdhill