Despite Kerckhoffs’s principle, proprietary or otherwise secret cryptographic algorithms are still used in real life. For security and efficiency reasons a common design practice simply modifies some parameters of widely used and well studied encryption standards. In this paper, we investigate the feasibility of reverse engineering the secret specifications of an AES-like block cipher by a FIRE attack based on Ineffective Fault Analysis (IFA) or by SCARE techniques based on two models of collision power analysis. In the considered fault or observational models, we demonstrate that an adversary who does not know the secret key can recover the full set of secret parameters of an AES-like software implementation and, in some models, even if it is protected by common Boolean masking and shuffling of independent operations. We thereby intend to demonstrate that protecting the implementation of such AES-like function is not an option even if its specifications are not public.