Sirvi Autor "Esslinger, Bernhard" järgi

Nüüd näidatakse 1 - 2 2

Cryptanalysis of Hagelin M-209 Cipher Machine with Artificial Neural Networks: A Known-Plaintext Attack
(Tartu University Library, 2024) Mikhalev, Vasily; Kopal, Nils; Esslinger, Bernhard; Lampesberger, Harald; Hermann, Eckehard; Waldispühl, Michelle; Megyesi, Beáta
This paper introduces a machine learning (ML) approach for cryptanalysis of the ciphermachine Hagelin M-2091. For recovering the part of the secret key, represented by the wheel pins, we use Artificial Neural Networks (ANN) which take as input the pseudo-random displacement values generated by the internal mechanism of the machine. The displacement values can be easily obtained when ciphertext and plaintext are known. In particular, we are using several distinct ANNs, each recovering exactly one pin. Thus, to recover all the 131 pins, we utilize 131 model seach solving a binary classification problem. By experimenting with various ANN architectures and ciphertext lengths, ranging from 52 to 200 characters, we identified an ANN architecture that outperforms others in accuracy. This model, inspired by the architecture by Gohr used for attacking modern ciphers, achieved the following accuracies in recovering the pins of the first wheel of the machine: approximately 71% for 52-characters sequences, 88% for 104-characters, 96% for 200-characters. The first wheel has the largest size and hence represents the most complicated case. For the other wheels, these accuracies are slightly higher. To the best of our knowledge, this is the first time when ANNs are used in a key-recovery attack against such machines.
Enhancing Classical Cipher Type Detection: Prompt Engineering with Common LLMs versus Usage of Custom AI Models
(Tartu University Library, 2025) Bastian, Maik; Esslinger, Bernhard; Hermann, Eckehard; Kopal, Nils; Lampesberger, Harald; Antal, Eugen; Marák, Pavol
In the field of cryptography, identifying the type of cipher used in an encrypted message is crucial to effective cryptanalysis. Thus far, from a machine learning perspective, this classification problem has been tackled using specifically designed models, such as the Neural Cipher Identifier (NCID), which require data generation and model training capabilities. The recent advent of Large Language Models (LLMs) raises the following question: Can this classification problem be approached more effectively through prompt engineering? This paper explores various generic strategies for prompt engineering, such as chain-of-thought and in-context learning, by evaluating thousands of generated prompts for classical ciphers using open-source LLMs (on an Nvidia DGX system) and ChatGPT (via a browser interface and API). The classification accuracies achieved through these prompting techniques are compared with those obtained by NCID. Although our findings indicate that NCID still significantly outperforms the use of LLMs for cipher-type detection, the latter offers a more accessible approach to cryptography tasks. Both methods can benefit from domain-specific knowledge in cryptanalysis, highlighting the importance of expert input in improving initial classifications and handling complex cipher types.