GAN-inspired Intelligent Jamming and Anti-jamming Strategy for Semantic Communication Systems
Rui Tang, Dahua Gao, Minxi Yang, Tao Guo, Huihui Wu, Guangming Shi
Abstract
The emerging semantic communication is currently receiving extensive attention. It focuses on semantic information transmission rather than the conventional symbol or bit transmission regardless of content. The semantic-level transmission in semantic communication brings new challenges for jamming and anti-jamming. How to attack and defend during semantic information transmission is the essence of jamming and anti-jamming in this communication paradigm. In this paper, we proposed a new intelligent jamming and anti-jamming framework to analyze and promote the security of semantic communication. In the proposed framework, the semantic jammer attempts to make the decoding result and the sent content semantically different. In contrast, the receiver tries to decode the semantic information correctly under the semantic jamming attack. To optimize the semantic jammer and receiver, we build a game model between them in the intelligent jamming and anti-jamming framework. And a game strategy like the generative adversarial network (GAN) is creatively introduced in the proposed framework. Thus, a more robust semantic communication system is built. We apply the proposed framework and strategy to text semantic communication. Multiple experiments were performed to verify the effectiveness of the proposed approaches.