Abstract: The most urgent problem to be solved in the prevention and attack of online fraud is how to accurately locate fraud websites from massive Internet resources. Therefore, we propose an active detection method for scam websites. By mining the footprint information left by the developers on infrastructures such as DNS resolution, cloud services, we could infer other potential scam websites based on a small set of known scam websites. We used the structural information and semantic information of the scam website to accurately identify the suspected scam website. Experiments show that the proposed active detection method can effectively mine suspected scam websites and identify them with high accuracy.