This talk addresses the issue of miniaturized cryptographic primitives, with particular 
focus on block ciphers. Examples include mini-versions of IDEA and the AES, which are 
straightforward to derive because they operate word wise. Mini-ciphers are often used as 
testbeds for new attacks, in order to collect data such as success rate and to corroborate 
theoretical predictions. An implicit assumption of this approach is that results on 
mini-ciphers shall equally apply to the original cipher. But, is it always the case? In 
this talk, I will present some experiments on higher-order square attacks applied 
to mini-AES versions.