Adversarial deepfake defence, production malware detection attacks, GPU-based ML authentication, multi-agent vulnerability detection, and training data scaling for adversarial robustness