Defending Collaborative Filtering Recommenders via Robustness-Aware Edge Reweighting

User-based collaborative filtering (CF) relies on a user–user similarity graph, which makes it vulnerable to profile injection (shilling) attacks that manipulate neighborhood relations to promote (push) or demote (nuke) target items. In this work, we propose an adversarial robustness–based edge reweighting scheme to defend CF. Specifically, the framework first assesses adversarial robustness to characterize edge-level reliability, thereby identifying non-robust edges that are likely to mislead the recommender under attacks and robust edges that remain stable. It then reweights edges according to their robustness to attenuate the influence of non-robust edges when making predictions. Experiments on benchmark datasets demonstrate the effectiveness of the proposed method.