This repository contains our implementations for UBA(an Uplift-guided Budget Allocation Framework) and various shilling attack methods including Leg-UP, AIA, WGAN, Random Attack, Average Attack, Segment Attack and Bandwagon Attack.
conda env create -f environment.yml
conda activate uba
You can download ML-1M, Yelp, Amazon
- Attack LightGCN model with baseline attacker Leg-UP
python -u main.py --data_set ml1m --model lgn --attacker_list AUSHplus --attack_num 300 --oneitem 3116 --allseed 2023 --way 1
- Attack mf model with UBA attacker AIA
python -u main.py --data_set ml1m --model mf --attacker_list AIA --attack_num 300 --oneitem 3116 --allseed 2023 --way 3
python DPA.py
For more example and parameters, please reference the example.sh .
Continue updating...