Binary fingerprinting codes with a length of theoretically minimum order were proposed by Tardos, and the traceability has been estimated under the well-known marking assumption. In this paper, we estimate the traceability and the false-positive probability of the fingerprinting code over AWGN channel, and propose a new accusation algorithm to catch more colluders with less innocent users. The design of our algorithm is based on the symmetric accusation algorithm proposed by Skorie et al. that focuses on the characteristic of the p.d.f. of the correlation scores. The proposed algorithm first estimates the strength of noise added to the code, and then calculates the specific correlation scores among candidate codewords using the characteristic of the noisy channel. The scores are finally classified into guilty and innocent by the threshold obtained from the p.d.f. The performance of the proposed tracing algorithm is evaluated by Monte Carlo simulation.