Graph Universal Adversarial Attacks: A Few Bad Actors Ruin Graph Learning ModelsXiao ZangYi Xieet al.2021IJCAI 2021