Abstract. In Wang’s attack, message modifications allow to deterministically satisfy certain sufficient conditions to find collisions efficiently. Unfortunately, message modifications significantly change the messages and one has little control over the colliding blocks. In this paper, we show how to choose some part of the messages which collide. Consequently, we break a security countermeasure proposed by Szydlo and Yin at CT-RSA ’06, where they added a fixed padding at the end of each block. Furthermore, we also apply this technique to partially recover the passwords in the Authentication Protocol of the Post Office Protocol (POP). This shows that collision attacks can be used to attack real protocols, which means that finding collisions is a real threat. Key words: Hash function, MD4, MD5, Wang, message modification for meaningful collisions, APOP security