Review of genetic algorithm for the example of guessing password by MD5-hashThe simplest method of password guessing by md5-hash consist in iterating through all possible options for passphrases and their comparison with md5- hashes heshom desired password. For this is a dictionary of all possible characters used in the password , and are made out of it all possible passphrases. The main disadvantage of this algorithm is that it is necessary to go through all the possible combinations of passphrases and it is unknown at which step the md5-hash of another combination of characters matches the md5-hash of the required passphrase. To reduce the time of selection of the passphrase you can use other methods, such as genetic algorithm. The essence of the genetic algorithm is as follows: given an initial random sample of passphrases - population size N phrases, with each phrase encoded in the genome - a chain consisting of genes. Genome is a binary code - a one-dimensional array of fixed-length or dynamic, where the element of the array - the gene takes the value 1 or 0. Thus, the initial population is generated randomly from the vocabulary - the set of all possible symbols used in the password, and presents a set of genomes. The first step of the algorithm is cross genomes - crossing, this randomly selected population of two genomes is selected randomly dividing point genome into two or more parts and then an exchange genes between the two genomes. Methods of gene exchange can be varied - for example, genes, reaching to the point of division of the first genome, genes are in place in the second genome, reaching to the point of division. Possible that the chosen few points of division of the genome - then there is an exchange - chain genes in a random or selected order. The crossing obtain new genomes - the descendants of a new set of genes, and in deciphering them - the new passphrase. The second step of the algorithm is a mutation - selective change genes in the genome. % Is subjected to a mutation in the population of genomes. This means that randomly selects a % of the genome of the new population consisting of genomes ancestor and descendant genome, produced in a previous stage and a change randomly selected genes opposite (1 to 0, 0 to 1 ) with a certain probability, for example 10%. The probability of mutation is selected in the settings of the algorithm. In the third step of the algorithm is a natural selection - truncated genomes do not satisfy the conditions. For that a survival function, which calculates the survival rate of the genome in the population. On the example of password guessing by md5-hash, it could be the number of characters matched in the hash and hash a passphrase desired password. At the selection stage, left N genomes of the new population, the most satisfying survival. Further there is a crossover, mutation and selection of a new population. The algorithm terminates when we find the desired passphrase, i.e. matched all the characters md5-hash a passphrase with symbols md5-hash of the password, respectively. The terms also stop the algorithm can be after a certain time to find a phrase or passage of a certain number of generations the algorithm. As a result of the genetic algorithm is a search password gradual approximation (identity) - increase in the number of matching symbols on the appropriate places in the md5-hash passphrases and md5- hash of the search phrase. Under certain conditions: the correct compilation of the population size, the method of crossing-over (crossover) and the percentage of mutations, the efficiency of the method by guessing the password md5-hash increases, and the search time is reduced.
|