---

Cryptography, the art and science of creating secret codes, and cryptanalysis, the art and science of breaking secret codes, underwent a similar and parallel course during history. Both fields evolved from manual encryption methods and manual codebreaking techniques, to cipher machines and codebreaking machines in the first half of the 20th century, and finally to computerbased encryption and cryptanalysis from the second half of the 20th century. However, despite the advent of modern computing technology, some of the more challenging classical cipher systems and machines have not yet been successfully cryptanalyzed. For others, cryptanalytic methods exist, but only for special and advantageous cases, such as when large amounts of ciphertext are available.

Starting from the 1990s, local search metaheuristics such as hill climbing, genetic algorithms, and simulated annealing have been employed, and in some cases, successfully, for the cryptanalysis of several classical ciphers. In most cases, however, results were mixed, and the application of such methods rather limited in their scope and performance.

In this work, a robust framework and methodology for the cryptanalysis of classical ciphers using local search metaheuristics, mainly hill climbing and simulated annealing, is described. In an extensive set of case studies conducted as part of this research, this new methodology has been validated and demonstrated as highly effective for the cryptanalysis of several challenging cipher systems and machines, which could not be effectively cryptanalyzed before, and with drastic improvements compared to previously published methods. This work also led to the decipherment of original encrypted messages from WWI, and to the solution, for the first time, of several public cryptographic challenges.

---