There are some preliminary results of the one-pixel attack performed on Cifar10 in the repo, but it is not quite as comprehensive as seen in https://arxiv.org/abs/1710.08864. It would be nice to not only replicate the experiments but also match (or surpass) their metrics.