NeurIPS 2020

Adversarial Attacks on Deep Graph Matching

Meta Review

The reviewers reached a concensus of accepting the paper. The authors propose novel and sound methods, supported with appropriate theoretical analysis, to address the problem of attacking graph matching, which is considered by the reviewers as a new problem. The reviewers are satisfied with the settings of empirical evaluation and experimental results. The authors are encouraged to carefully revise their paper to address some remaining concerns of the reviewers.