The strength of key-expansion algorithm is one of the important impacts on the security of block ciphers.In practical use,recovering the master key from a subkey is a threat to the reliability of cipher algorithm.However,there has not been sufficient research dedicated to analyzing its security strength.Key-expansion algorithm with truncation of lightweight block ciphers is focused on in this pa-per.According to this algorithm,the master key bits are classified into cut-off bits and remained bits.The master keys are recovered from last-round subkeys to explore the strength of these key-expansion algorithms using deep learning.A fully connected neural network is trained and its recovery accuracy of remained bits is higher than that of cut-off bits.Most bits of the master keys are recovered by the convolutional neural network with an accuracy above 0.7.Based on it,the strength of key-expansion al-gorithms with truncation of lightweight block ciphers PRESENT,GIFT,and MIBS is further analyzed.
维普
万方数据
