Adversarial-AES

※This code is under development.

This is an AES model using a domain adversarial neural network.

Requirements

• python = 3.8.10
• tensorflow = 2.9.2
• numpy = 1.21.6
• pandas = 1.3.5
• nltk = 3.7

Training

If you want to use word series as an input, you should download glove.6B.50d.txt from https://nlp.stanford.edu/projects/glove/ and place in the embeddings folder.

Also, download mnistm.h5 from https://github.com/sghoshjr/tf-dann/releases/download/v1.0.0/mnistm.h5 if you want to try domain adaptation on the MNIST dataset and place in the Datasets/MNIST_M folder.

• MNIST/MNIST_M domain adaptation

  • Run train_sampleDANN.py.

• ASAP domain adaptation

  • Run the train_Smodel.py script. you can chose options: --train_mode domain-adaptation/source, --with_features yes/no, --input_seq words/pos.