We propose a novel membership inference attack based on predicted labels, which is the first black‐box membership inference attack against machine unlearning in the ...