New research found a way to both improve the accuracy of deepfake detection algorithms while also enhancing fairness.