A challenge-response approach can effectively detect real-time deepfakes by exploiting inherent limitations of deepfake generation pipelines.
