In order to improve the data embedding capacity of coverless information hiding algorithm, scholars have done a lot of research work to promote its practical research after it was proposed. By researching the user’s behavioralShabits of several social platforms, for improving the capacity of data embedding and the efficiency of communication, the paper introduced a semi-structured coverless information hiding algorithm based on the behavioralShabits of social platforms.Some behavioral habits in social platforms are extracted through deep learning and data mining technology, and small icons are spliced together based on behavioral habits to implement the graphical representation of secret message and achieve the transmission of secret message in the algorithm.By using the convolutional neural network to train and classify, the icons in the icon library can be recognized, a one-to-one mapping is built between icons and secret messages. In order to improve the robustness of the algorithm, some icons processed by various attack methods are introduced as interference samples in the training set. The experimental results can show that the algorithm has good anti-attack ability and the hiding capacity can be improved, the algorithm can be used in covert communication