Unsupervised Cipher Cracking Using Discrete GANs

TL;DR

This paper introduces CipherGAN, a novel unsupervised GAN-based architecture capable of cracking classical ciphers like shift and Vigenere on large vocabularies, by adapting CycleGAN for discrete data.

Contribution

CipherGAN is the first GAN-based method to effectively crack classical ciphers on large vocabularies using unpaired data, with stable training and discrete data compatibility.

Findings

Successfully cracks shift and Vigenere ciphers with high fidelity

Handles vocabularies larger than previous methods

Avoids uninformative discrimination in GAN training

Abstract

This work details CipherGAN, an architecture inspired by CycleGAN used for inferring the underlying cipher mapping given banks of unpaired ciphertext and plaintext. We demonstrate that CipherGAN is capable of cracking language data enciphered using shift and Vigenere ciphers to a high degree of fidelity and for vocabularies much larger than previously achieved. We present how CycleGAN can be made compatible with discrete data and train in a stable way. We then prove that the technique used in CipherGAN avoids the common problem of uninformative discrimination associated with GANs applied to discrete data.