Skip to content

Tensorflow implementation of TrialAttack (Triple Adversarial Learning for Influence based Poisoning Attack in Recommender Systems. KDD 2021)

Notifications You must be signed in to change notification settings

Daftstone/TrialAttack

Repository files navigation

Triple Adversarial Learning for Influence based Poisoning Attack in Recommender Systems

This project is for the paper: Triple Adversarial Learning for Influence based Poisoning Attack in Recommender Systems, Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 2021: 1830-1840.

The code was developed on Python 3.6 and tensorflow 1.14.0.

Usage

run generate_fake.py

usage: python generate_fake.py [--dataset DATA_NAME] [--gpu GPU_ID]
[--epochs EPOCHS] [--data_size DATA_SIZE] [--target_index TARGET_ITEMS]

optional arguments:
  --dataset DATA_NAME
                        Supported: filmtrust, ml-100k, ml-1m.
  --gpu GPU_ID
                        GPU ID, default is 0.
  --epochs EPOCHS
                        Training epochs.
  --data_size DATA_SIZE
                        The data available to the attacker.
  --target_index TARGET_ITEMS
                        The index of predefined target item list: 0, 1 for ml-100k, 2,3 for ml-1m, 4,5 for filmtrust, 6,7 for yelp.

Example.

python generate_fake.py --dataset ml-100k --gpu 0 --target_index 0

About

Tensorflow implementation of TrialAttack (Triple Adversarial Learning for Influence based Poisoning Attack in Recommender Systems. KDD 2021)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages