We propose a matching algorithm secure against the wolf attack in one-to-one biometric authentications. Our proposal algorithm first executes a one-to-all comparison using a conventional measure such as the Hamming distance and then decides whether the prover is truly an authorized user by the number of unauthorized users who are incorrectly accepted. We calculate upper bounds of FAR, FRR and WAP (the wolf attack probability) in the proposal authentication algorithm and show the trade-off between an upper bound of WAP and that of FRR. We also show that our proposal algorithm can achieve lower FAR, FRR and WAP simultaneously. 1