This project is based on:Greydanus, S. (2017). Learning the enigma with recurrent neural networks. arXiv preprint arXiv:1708.07576. The paper has make an excellent exploration on learning the cipher by recurrent neural network. Showing that the neural network has ability to learn the cipher. Our project would try to step further from the paper, especially of the part that regard to Enigma. In future, we would try different architecture for keyphrase construct task(seq2seq model), or extend it to reconstruct the other configs of Enigma(Rotor type, Ring setting, plugboard, reflector, etc)