With the vigorous development of the multi-chain ecosystem, cross-chain bridges, as the infrastructure supporting blockchain interoperability, have become increasingly important. However, in recent years, there have been dozens of attacks on cross-chain bridges, resulting in billions of dollars in losses. Nevertheless, existing research on detecting cross-chain attacks has limitations. In this paper, we have studied real attack incidents that have occurred in recent years and, based on this, documented three types of cross-chain attacks. To detect these types of attack, we propose CIFGViewer, a cross-chain attack detection tool from the perspective of cross-chain bridge access control. Specifically, CIFGViewer constructs cross-chain information flow graph (CIFGs) based on transaction data, mines potential account-permission assignment relationships from the CIFGs to generate a cross-chain access control model, and finally combines security patterns to detect attack transactions. To evaluate the effectiveness of CIFGViewer, we constructed a dataset containing 151 attack transactions from real attack incidents. The evaluation results show that CIFGViewer has an accuracy rate of 96% in detecting cross-chain attacks. In addition, our tool identifies 54 potential attack transactions among 7,756 real cross-chain transactions.