The internet auction is one of the most popular transaction ways in electronic commerce. In this paper, we develop a method which contains two steps to detect fraud bidders for one auction product. First, we compare the bidders ever buying products with bidding products in the transaction data. If the results do not conform the assigned condition, the detecting process has to use the method of second step. We let the auctioning product as the center of cluster, and let each other transaction data are grouped with the center to which it contains the auctioning product in ever buying products. Then we compute the association items from the c1uster.We can consider the bidders whether fraud or not according to the association items for the auctioning product.