v1.2.0

Trained on 27,977 samples, these weights have achieved an accuracy in classification of 87.0000% (evaluation). This model expects the input to be the full binary represented as a 1024x1024 image.

The accuracy figure comes from a set of samples unique from the training dataset. See the README for more detailed statistics.

Full Changelog: v1.1.1...v1.2.0

v1.1.1

Trained on 13.2k samples, these weights have achieved an accuracy in classification of 82.5000%. This model expects the input to be the full binary represented as a 1024x1024 image.

The accuracy figure comes from a set of samples unique from the training dataset. See the README for more details.

Full Changelog: v1.1.0...v1.1.1

v1.1.0

Trained on 2k samples, these weights have achieved an accuracy in classification of 75.0000%. This model expects the input to be the full binary and not the feature-extracted .text section of the PE file.

The accuracy figure comes from a set of samples unique from the training dataset. With that in mind, the weights of the previous model have been retested and achieved an accuracy of 57.500%.

Full Changelog: v1.0.0...v1.1.0

v1.0.0

Initial Release. Trained on ~2k samples, these weights have achieved an accuracy in classification of 82.4121%.