In 2013, Borghoff \emph{et al}. introduced a slender-set linear
cryptanalysis on PRESENT-like ciphers with key-dependent secret
S-boxes. In this paper, we propose an improved slender-set linear
attack to PRESENT-like ciphers with secret S-boxes. We investigate
three new cryptanalytic techniques, and use them to recover the
secret S-boxes efficiently. Our first new idea is that we propose a
new technique to support consistency of partitions of the input to
the secret S-boxes. Our second new technique is that we present a
more efficient method to recover the coordinate functions of secret
S-boxes based on more information than that of Borghoff's attack.
The third new technique is that we propose a method of constructing
all correct coordinate function of secret S-boxes by pruning search
algorithm. In particular, we implemented a successful linear attack
on the full round Maya in practice. In our experiments, the correct
S-box can be recovered with $2^{36}$ known plaintexts, $2^{18.9}$
time complexity and negligible memory complexity at a success rate
of 87.5\%. Our attack is the improvement and sequel of Borghoff's
work on PRESENT-like cipher with secret S-boxes.
↧