Large-scale Testing Global Optimization Methods with Black-box Adversarial Attacks