Kahden jonosalausmenetelmän lineaarinen kryptoanalyysi
Hakala, Risto · Aaltodoc (Aalto University) · 2007
Stream ciphers are symmetric encryption primitives, which are used to ensure confidentiality of messages in digital communications.Compared to block ciphers, stream ciphers are often more efficient and allow a more compact implementation-they are suitable especially for telecommunica tion applications.The security of stream ciphers has not been on the same level with the most secure block ciphers, however.This is why the design and analysis of stream ciphers has started to receive even more attention.In this thesis, a survey of linear cryptanalysis of shift register-based stream ciphers is performed, and attacks on two recently developed stream ciphers, SOBER-128 and Shannon, are presented.The attacks are linear distin guishing attacks, which aim at distinguishing a keystream from a truly random sequence using linear cryptanalytic techniques.A linear distin guishing attack is based on a linear transformation, which is able to detect statistical bias in the keystream.The transformation that is used in the distinguishing attack is formed by approximating the nonlinear parts in the cipher with linear functions.In order to find a good transformation, we present a new technique for analyzing linear approximations of T-functions efficiently.This technique is used to construct the distinguishing attack on SOBER-128, in which case the attack also gives information about the secret constant that is used in SOBER-128.The distinguishing attack on Shannon is based on a multidimensional linear transformation.A notable benefit is gained from using a multidimensional transformation instead of a one-dimensional transformation.