In this paper, we propose a fingerprinting solution to protect valuable numeric relational data from illegal duplications and redistributions. We introduce a twice-embedding scheme. In the first embedding process, we embed a unique fingerprint to identify each recipient to whom the relational data is distributed. The embedding process is controlled by a secret key. Meanwhile, the fingerprint can be detected using the same secret key to prove ownership at a numerical confidence level. The second embedding process is designed for verifying the extracted fingerprint and giving a numerical confidence level. Thus, once a suspect copy is found, numerical confidence level can be provided both to identify the owner and the illegal distributor. The experiment shows that our solution is effective and robust to various attacks.