Abstract: Current optical coherence tomography(OCT) based fingerprint anti-counterfeiting methods are mainly through the extraction of sweat pores,subcutaneous sweat glands and internal fingerprints,which depend on the extraction accuracy and contain redundant anti-counterfeiting information.An OCT fingerprint anti-counterfeiting method based on the attention residual network is proposed,where only a small amount of data is required to identify the authenticity of the fingerprint.Firstly,the residual network is incorporated with the convolutional block attention module as well as the long and short skip connections to better focus the network attention on the subcutaneous tissue information.Secondly,random sampling is performed on the volume data of the discriminating finger,and the local feature patches are extracted.In the last,these local feature patches are classified by the attention residual network to realize the identification of the true and fake fingerprint.Experiments are carried out using the Keras framework and on the self-established true and fake fingerprint database.The proposed method obtains a higher anti-counterfeiting accuracy rate than other networks,which proves the superiority of the proposed method.The cross-device and cross-sample experiment on self-established database and Shenzhen University OCT fingerprint database has further verified the accuracy and robustness of the proposed method.