The invention belongs to the field of information technology security, and particularly relates to a CNN and RNN fusion model-based network heterogeneous concurrent steganalysis channel detection method. The method comprises the following steps of: extracting detection features of the multi-dimensional concurrent covert channel based on deep learning; learning the extracted detection characteristics; constructing a multi-dimensional concurrent covert channel detector and carrying out detection; According to the method, the defects of large feature dimension and incapability of expressing multi-layer data complex association caused by artificial design features based on experience or heuristic knowledge are avoided, and the detection of multi-dimensional concurrent covert communication of the network streaming media is realized by utilizing an automatic representation method for automatically mining feature association based on deep learning.