# Unsupervised Cipher Cracking Using Discrete GANs

@article{Gomez2018UnsupervisedCC, title={Unsupervised Cipher Cracking Using Discrete GANs}, author={Aidan N. Gomez and Sicong Huang and Ivan Zhang and Bryan M. Li and Muhammad Osama and Lukasz Kaiser}, journal={ArXiv}, year={2018}, volume={abs/1801.04883} }

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…

## 45 Citations

Neural Cryptanalysis of Classical Ciphers

- Computer Science, MathematicsICTCS
- 2018

Artificial neural networks are applied to automatically “assist” cryptanalysts into exploiting cipher weaknesses and provide the first ciphertext-only attack on substitution ciphers based on neural networks.

Output Prediction Attacks on SPN Block Ciphers using Deep Learning

- Computer Science, MathematicsIACR Cryptol. ePrint Arch.
- 2021

This paper focuses on toy SPN block ciphers (small PRESENT and small AES) and proposes deep learningbased output prediction attacks that realize output predictions that are much stronger than distinguishing attacks, and demonstrates that swapping the order of components or replacement components affects the success probabilities of the attacks.

Assessing Lightweight Block Cipher Security using Linear and Nonlinear Machine Learning Classifiers

- Computer Science, Mathematics
- 2021

This paper trains six machine learning classifiers (linear and nonlinear) to perform the security prediction task using a dataset generated from a small-scale generalized Feistel structure (GFS) cipher as a proof-of-concept and applies the proposed approach to a full-scale lightweight GFS block cipher, TWINE.

Hidden Markov Models for Vigenère Cryptanalysis

- Computer Science, MathematicsHistoCrypt
- 2018

This paper shows that hidden Markov models are also applicable to the cryptanalysis of the well-known Vigenère cipher, and argues that the model generated by an HMM is considerably more informative than that produced by a GAN.

Output Prediction Attacks on Block Ciphers using Deep Learning∗

- Computer Science, Mathematics
- 2021

This paper focuses on two toy SPN block ciphers and one toy Feistel block cipher and proposes deep learning-based output prediction attacks and demonstrates the following: attacks work against a similar number of rounds attacked by linear/differential cryptanalysis, their attacks realize output predictions that are much stronger than distinguishing attacks, and swapping the component order or replacement components affects the success probabilities of the proposed attacks.

Deep Learning-Based Cryptanalysis of Lightweight Block Ciphers

- Computer Science, MathematicsSecur. Commun. Networks
- 2020

The proposed generic cryptanalysis model based on deep learning (DL), where the model tries to find the key of block ciphers from known plaintext-ciphertext pairs, shows the feasibility and indicates that the DL technology can be a useful tool for the cryptanalysis of blockciphers when the keyspace is restricted.

Assessing Block Cipher Security using Linear and Nonlinear Machine Learning Models

- Computer Science, MathematicsIACR Cryptol. ePrint Arch.
- 2020

Findings show that nonlinear classifiers outperform linear classifiers for the prediction task due to the nonlinear nature of block ciphers, and indicate the feasibility of using the proposed approach in assessing block cipher security or as machine learning distinguishers.

Lightweight Block Cipher Security Evaluation Based on Machine Learning Classifiers and Active S-Boxes

- Computer Science, MathematicsIEEE Access
- 2021

This work investigates the capability of linear and nonlinear machine learning classifiers in evaluating block cipher security by training the best performing nonlinear classifiers using data from other similar ciphers, and showcases the feasibility of using simple machine learningclassifiers as a security evaluation tool to assess block cipherSecurity.

Linear Attack on Round-Reduced DES Using Deep Learning

- Computer Science, MathematicsESORICS
- 2020

This work uses deep learning networks to achieve linear attack on DES with plain-cipher pairs for the first time, and indicates that trained neural networks can effectively learn algorithmic representations of the XOR distributions of given linear expression on DES.

Convolutional neural network based cipher algorithm attack scheme

- Computer Science
- 2021

The experimental results reflect that the model designed based on convolutional neural network is more sensitive to the results of complex password decryption, and the model is able to reduce the distance in the space between the original plaintext and the decrypted plaintext.

## References

SHOWING 1-10 OF 42 REFERENCES

Solving Substitution Ciphers

- Computer Science
- 2007

Experimental results show a median of 94% cipher letters correctly decoded, which is typically good enough for an unskilled human to finish decoding the cipher with minimal additional effort.

Decipherment of Substitution Cipher using Enhanced Probability Distribution

- Computer Science, Mathematics
- 2010

The retrieved efficiency of cipher text only attack on samples of English, Hindi Telugu, Kannada is presented and enhanced probability distribution function is adopted while evaluating the enhanced probabilities of message units.

A cryptanalytic attack on Vigenère cipher using genetic algorithm

- Computer Science, Mathematics2011 IEEE Conference on Open Systems
- 2011

The cryptanalysis of a poly alphabetic substitution cipher (Vigenère cipher) by applying genetic algorithms is presented and the applicability of genetic algorithms for searching the key space of encryption scheme is studied.

The Automated Cryptanalysis of Substitution Ciphers

- Computer ScienceCryptologia
- 1986

Expert systems programmed in artificial intelligence languages can break simple substitution ciphers by capturing and employing knowledge about the relative frequencies of single letters and certain…

Genetic Algorithm and Tabu Search Attack on the Mono-Alphabetic Subsitution Cipher in Adhoc Networks

- Computer Science
- 2007

This study presents a cryptanalysis method based on Genetic Algorithm and Tabu Search to break a Mono-Alphabetic Substitution Cipher in Adhoc networks and compares and analyzed the performance of these algorithms in automated attacks on Mono-alphabetic Substitutes Cipher.

Applying Genetic Algorithms for Searching Key-Space of Polyalphabetic Substitution Ciphers

- Computer ScienceInt. Arab J. Inf. Technol.
- 2008

The Cryptanalysis of polyalphabetic by applying genetic algorithm by applying Genetic algorithm is presented and the applicability of Genetic algorithms for searching the key space of encryption scheme is studied.

The Copiale Cipher

- Computer Science, MathematicsBUCC@ACL
- 2011

The Copiale cipher is a 105-page enciphered book dated 1866 and the features of the book and the method by which it is deciphered are described.

Unsupervised Analysis for Decipherment Problems

- Computer ScienceACL
- 2006

Techniques for understanding errors and significantly increasing performance in natural language decipherment problems using unsupervised learning are described.

How to Train Your DRAGAN

- Computer ScienceArXiv
- 2017

This paper introduces regret minimization as a technique to reach equilibrium in games and uses this to justify the success of simultaneous GD in GANs and develops an algorithm called DRAGAN that is fast, simple to implement and achieves competitive performance in a stable fashion.

Boundary-Seeking Generative Adversarial Networks

- Computer ScienceICLR 2017
- 2017

This work introduces a method for training GANs with discrete data that uses the estimated difference measure from the discriminator to compute importance weights for generated samples, thus providing a policy gradient for training the generator.