Decoding the Enigma with Recurrent Neural Networks

Decoding the Enigma is interesting, but these days it has no practical use. Modern encryption uses public-key factoring algorithms such as RSA. RSA is a different beast from the Enigma, but in theory we could also decode it with deep learning. In practice, this is difficult because RSA uses modulus and multiplication of large integers. These operations are difficult to approximate with RNNs. We need further algorithmic advances in deep learning like the Neural GPU or the Differential Neural Computer to make

Source: Decoding the Enigma with Recurrent Neural Networks