Conventional cryptographic algorithms are not sufficient to protect secret keys and data in white‐box environments, where an attacker has full visibility and control over an executing software code. For this reason, cryptographic algorithms have been redesigned to be resistant to white‐box attacks. The first white‐box AES (WB‐AES) implementation was thought to provide reliable security in that all brute force attacks are infeasible even in white‐box environments; however, this proved not to be the case. In particular, Billet and others presented a cryptanalysis of WB‐AES with 230 time complexity, and Michiels and others generalized it for all substitution‐linear transformation ciphers. Recently, a collision‐based cryptanalysis was also reported. In this paper, we revisit Chow and others’ first WB‐AES implementation and present a conditional re‐encoding method for cryptanalysis protection. The experimental results show that there is approximately a 57% increase in the memory requirement and a 20% increase in execution speed.