Conceptual

Security and Privacy Techniques for Semantic Communication Systems

How semantic-communication systems - which transmit deep-learning-extracted meaning rather than raw bits - can be attacked and defended across their life cycle. Students learn the threat surface at model training, model transfer, and semantic transmission (poisoning, backdoor, adversarial, and inference attacks) and the defense toolkit spanning robust and adversarial training, differential privacy, cryptography, blockchain, model compression, and physical-layer security.