In this work, we extend the FGSM method proposing multistep adversarial perturbation (MSAP) procedures to study the recommenders’ robustness under powerful methods. Letting fixed the perturbation magnitude, we illustrate that MSAP is much more harmful than FGSM in corrupting the recommendation performance of BPR-MF.
security
collaborative-filtering
recommender-system
aml
adversarial-machine-learning
fgsm
perturbations
msap
fast-gradient-sign-attack
bayesian-per
-
Updated
Dec 8, 2022 - Python