Efficient Project Gradient Descent for Ensemble Adversarial Attack
This work addresses the need for more effective adversarial attacks in machine learning security, particularly for ensemble models, though it appears incremental as it builds on existing PGD methods.
The paper tackles the problem of generating adversarial examples with smaller perturbations for ensemble models by proposing an efficient modified Project Gradient Descent method that automatically adjusts ensemble weights and step size per iteration, achieving first place in the IJCAI19 Targeted Adversarial Attack competition.
Recent advances show that deep neural networks are not robust to deliberately crafted adversarial examples which many are generated by adding human imperceptible perturbation to clear input. Consider $l_2$ norms attacks, Project Gradient Descent (PGD) and the Carlini and Wagner (C\&W) attacks are the two main methods, where PGD control max perturbation for adversarial examples while C\&W approach treats perturbation as a regularization term optimized it with loss function together. If we carefully set parameters for any individual input, both methods become similar. In general, PGD attacks perform faster but obtains larger perturbation to find adversarial examples than the C\&W when fixing the parameters for all inputs. In this report, we propose an efficient modified PGD method for attacking ensemble models by automatically changing ensemble weights and step size per iteration per input. This method generates smaller perturbation adversarial examples than PGD method while remains efficient as compared to C\&W method. Our method won the first place in IJCAI19 Targeted Adversarial Attack competition.