Investigations in metaheuristic techniques with application to cryptology

Abstract

The objective of the research reported in this Ph.D. thesis is to investigate the effectiveness and efficiency of various metaheuristic techniques in the area of automated - cryptanalysis, cryptographic key and function generations. Automation of many processes related to cryptology removes the need for timeconsuming interaction of the human with the search process, therefore, the use of metaheuristics is desirable. For the development of the algorithms for solving some of the cryptology problems, mainly four metaheuristic techniques are utilized in this thesis that are as follows: genetic algorithm, particle swarm optimization, cuckoo search, and Cartesian genetic programming. In the case of cryptanalysis of classical ciphers, the n-gram statistics of the language are utilized by the metaheuristic techniques to determine the secret key in the large key space. In theoretical computer science the problem of determining the secret key in large key space falls under NP-complete problem. Since the genetic algorithm has been widely utilized for automated cryptanalysis of classical substitution and transposition ciphers, therefore, in this work, we have improved the existing genetic algorithm and also proposes a new technique based on the cuckoo search for automated cryptanalysis of classical substitution and transposition ciphers. The results indicate that the proposed cuckoo search strategy is highly efficient, highly effective and more successful technique when compared to the existing attacking techniques of the substitution cipher and the transposition cipher. During the past decade, considerable improved versions of real particle swarm optimization (PSO) have been proposed in the literature. However, only a few significant improved versions of binary PSO have been reported. For efficiently solving binary optimization problems, in this thesis, we propose an improved binary-PSO technique in which an improved idea for updating particles’ velocity is proposed. In order to escape from the local optimum, an on/off mutation strategy is also introduced between updates of particles’ velocity and particles’ position. To prove the effectiveness, the proposed technique is applied on solving two reduced knapsack cryptosystems. To assess the proposed technique, two relatively new methods, namely, novel binary-PSOand modified binary-PSO are also utilized. This research also proposes an improved genetic algorithm and a binary cuckoo search technique. The experimental results obtained through proposed techniques are analyzed statistically by conducting f-test and t-test. The f-test and t-test results indicate that the improved binary PSO strategy handles the considered automated cryptanalysis problem efficiently in terms of accuracy and convergence. In 2013, a genetic-based random key generator (GRKG) for generating the one-time password or pad (OTP) has been proposed in the literature which has certain limitations. In this research, two main characteristics (speed and randomness) of the GRKG method are significantly improved by presenting the IGRKG method (improved genetic-based random key generator method). The proposed IGRKG method generates an initial pad by using the linearcongruential generator and improves the randomness of the initial pad using genetic algorithm. The experimental results shows the superiority of the IGRKG over GRKG in terms of speed and randomness. Hereby we would like to mention that no prior experimental work has been presented which is directly related to the OTP key generation using various evolutionary algorithm. Therefore, this work can be considered as a guideline for future research. Recent years have also witnessed the significant increase in both the sidechannel attacks and defending mechanism on the cryptographic algorithms. We know that the differential power analysis (DPA) is the most powerful attack which belongs to the class of side channel attacks. In order to defend against DPA attacks, there is a growing demand for the construction of Boolean functions and vectorial Boolean functions. In this regard, we also develop effective algorithms based on evolutionary computing techniques. As a result, three 8-bit highly nonlinear balanced Boolean functions have been evolved in this work that has higher DPA resistance than the existing.