At the RFID Security Workshop 2007, Adi Shamir presented a new challenge-response protocol well suited for RFIDs, although based on the Rabin public-key cryptosystem. This protocol, which we call SQUASH-0, was using a linear mixing function which was subsequently withdrawn. Essentially, we mount an attack against SQUASH-0 with full window which could be used as a ``known random coins attack'' against Rabin-SAEP. We then extend it for SQUASH-0 with arbitrary window. We apply it with the proposed modulus $2^{1,277}-1$ to run a key recovery attack using $1,024$ chosen challenges. Since the security arguments equally apply to the final version of SQUASH and to SQUASH-0, we challenge the blame-game argument for the security of SQUASH. Nevertheless, our attacks are inefficient when using non-linear mixing so the security of SQUASH remains open.