Recent Methods for Cryptanalysis of Symmetric-key Cryptographic Algorithms (Recente Methoden voor de Cryptanalyse van Symmetrische-sleutel Cryptografische Algoritmen)

Cryptography is the art and science of secret communication. In the past it has been exclusively the occupation of the military. It is only during the last forty years that the study and practice of cryptography has reached the wide public. Nowadays, cryptography is not only actively studied in leading universities as part of their regular curriculum, but it is also widely used in our everyday lives. It protects our GSM communications and on-line financial transactions, our electronic health records and personal data. Internet services for which security is critical, such as online banking, electronic commerce, e-voting and the whole concept of the e-Government are utterly unimaginable without the necessary cryptographic mechanisms. In order for cryptography to serve its purposes well, secure and reliable cryptographic algorithms are necessary. The design of such algorithms is intimately linked to the ability to analyze and understand their properties. The latter are the subject of study of cryptanalysis. A cryptanalytic technique to a cryptographer is what the hammer and the anvil are to the blacksmith. With better tools higher art is accomplished. The goal of this thesis is to study new techniques for cryptanalysis of symmetric-key cryptographic algorithms. The first part of the thesis focuses on methods for cryptanalysis of ARX algorithms. These are algorithms based on the operations modular addition, bit rotation and XOR, collectively denoted as ARX. Many contemporary algorithms fall into this class. For example, the block ciphers TEA, XTEA and RC5, the stream cipher Salsa20, the hash functions MD4, MD5, SHA-1 and SHA-2 as well as two of the candidate proposals for the next generation cryptographic hash function standard SHA-3: the hash functions BLAKE and Skein. In this thesis we propose a general framework for the differential analysis of ARX algorithms. This framework is used to compute the probabilities with which differences propagate through the ARX operations. The accurate computation of these probabilities is critical for estimating the success of one