cryptic

***** Utility of each files *****

Several models have been trained for the task:

• Auto Encoder: This model aims at learning the representations of the binaries and then using them to classify.
• LSTM: This model uses the LSTM architecture, which takes each binary bit for processing and classifies.
• Dense: This is the vanilla neural network (ANN) for classification.

How to use:

• Note:
  • The data file, "TrainingData.csv" needs to be in the same directory as the python files.
  • Approaches have been made with fixed seed values for reproducibility.

For the Auto Encoder model:

• If you want to retrain it, run "auto_enc.py". This will save the best-trained encoder model as "best_encoder.h5".
• Next, run "encoded_inputs_lstm.py" to train the classification model on the encoded inputs. The model will be saved as "encoded_inputs_model.h5"

For the LSTM model:

• Run "lstm_tf.py" to train the classification model. The saved model will be named "best_classifier_2.h5"

For the Dense model:

• Run "dense_tf.py" to train the classification model. The saved model will be named "best_classifier_dense.h5"

The results in "Comparative_results.pdf" are evaluated on the validation set.

Packages

• python(3.10.13)
• tensorflow(2.14.0)

Contact

Please post a Github issue or contact ([email protected]) if you have any questions.