Building Undetectable Covert Channels Over Mobile Networks with Machine Learning

Covert channel is an important way to transmit covert message and implement covert communication through the network. However, the existing research on covert channel cannot meet the security requirements of covert communication in the complex mobile networks. There are problems such as low transmission capacity, insufficient adaptability to network complexity, and difficulty in countering the detection of covert channels by adversaries. In this paper, we preprocess video traffics over mobile network, and extract traffic features to build a target model. We analysis traffic data by machine learning method to improve the undetectability of the covert channel. Based on the characteristics of real-time interactive communication, gray code and interval block are employed to improve the robustness of covert communication in the complex network environment. A cover channel over VoLTE video traffic, which is based on video packet reordering supported by machine learning algorithms, is proposed to realize the awareness and confrontation of detection attacks on the network side. The covert channel is built over mobile network to ensure end-to-end reliable covert communication under complex network conditions.