Universal Distributional Decision-based Black-box Adversarial Attack with Reinforcement Learning. (arXiv:2211.08384v1 [cs.LG])

The vulnerability of the high-performance machine learning models implies a
security risk in applications with real-world consequences. Research on
adversarial attacks is beneficial in guiding the development of machine
learning models on the one hand and finding targeted defenses on the other.
However, most of the adversarial attacks today leverage the gradient or logit
information from the models to generate adversarial perturbation. Works in the
more realistic domain: decision-based attacks, which generate adversarial
perturbation solely based on observing the output …

adversarial attack box decision