It seems that algebraic attacks against cipher algorithms have not succeeded in practically breaking ciphers. The reason primarily is that the algebraic models of ciphers under with a known plaintext ciphertext pair and unknown key bits are too complex both in structure and number of variables. However if number of variables are brought down in a model to a sufficiently small number then it is possible to find structural weaknesses of equations or systematic algebraic heuristics of solution algorithms to find key bits. Although current algebraic models of AES have 9000+ variables and equations, these models contain mostly latent variables which can be eliminated algebraically offline. Hence in principle a 128 variable model of AES128 with a given plaintext ciphertext pair of blocks can be developed from an offline algebraic model. Should this not be a warning that AES can in principle be broken even with classical computers? Some agencies might already have such models.
Asked
Active
Viewed 53 times
0
these models contain mostly latent variables which can be eliminated algebraically offlinerequires a reference. You write in a way that there is an algebraic attack that breaks the AES. – kelalaka Nov 03 '19 at 07:54