Maxi-linear Deviation Searching Algorithm Based on Improved Pyramid Filter

To solve the problem of calculating the maximum linear deviation in block cipher algorithm linear analysis, the paper proposes a searching method based on improved pyramid filter of the maximum deviation. The ciphertext and its coefficient matrix is initialized as pyramid matrix, using ICP algorithm to revise and calculate the best match state of adjacent pyramid matrix from the high level to the low level that largely reduces the searching consuming and enhances the accuracy, the linear deviation value of round function transformation at the best fitting time is proved to be the maximum linear deviation by applying generalized Walsh spectrum. It determine the initial corresponding relation of iterative algorithm by proposing the initial rule corresponding distance, each round of transformation matrix would be assigned to the next round as the initial transformation matrix has also improved the matching speed of the ICP algorithm; the rationality and effectiveness of the new searching algorithm can be seen by comparing with the maximum linear deviation search algorithm put forward by M. Matsui.