With the help of mobile phones, the identification of forged packaging printed matter can be realized, and the active confirmation can be realized, so as to achieve more accurate identification of true and fake packaging printed matter. In the case of extremely lack of fake samples, a sample generation algorithm based on DCGAN memory enhancement was proposed to realize the amplification of fake samples, and the attention convolutional neural network was used to distinguish the authenticity of packaged printed matter defensively. Several groups of authentic packaging printed matter and a very small number of counterfeit packaging printed matter images were shot in four open scenes, which consisted of one kind of printing paper, two kinds of shooting light source and two kinds of shooting mobile phone. The sample of counterfeit packaging printed matter was generated based on DCGAN memory enhancement algorithm, and the algorithm research data set was established. The experimental results show that the discriminant accuracy of S-PA-CNN twin attention convolutional neural network based on amplified data set is more than 97%. The experimental results show that the proposed data amplification method can further improve the authenticity feature recognition ability of the network model, improve the fine-grained discrimination accuracy, and enhance the generalization ability.