RICC is a robust collective classification framework designed to identify Sybil accounts on online social networks. We observed that the classification results for adversarial Sybil accounts often significantly change when deploying a new training set different from the original training set. Leveraging this observation, RICC achieves robustness against state-of-the-art adversarial attacks by stabilizing classification results across different training sets randomly sampled in each round. For more details, please refer to our paper, "RICC: Robust Collective Classification of Sybil Accounts", which appeared in The Web Conference (WWW) 2023.
We implemented RICC in Python and tested on a machine running Ubuntu 20.04.5 LTS and Python 3.8. To get ready for running RICC, please install the dependencies by running the following commands.
$ git clone https://github.com/WSP-LAB/RICC.git
$ cd RICC
$ pip install -r requirements.txt
We provide the Enron and Facebook datasets in the dataset directory. These
datasets are from the Stanford Large Network Dataset Collection (SNAP). You can
access the original datasets from here.
For the Twitter_small and Twitter_large datasets, we refer the users to the following links.
Twitter_small dataset : https://success.cse.tamu.edu/releases/
Twitter_large dataset : http://wangbinghui.net/dataset.html
Please refer to this link for writing a configuration file.
To run RICC, execute the script RICC.py by passing the configuration file as
an argument.
$ cd RICC/script
$ python RICC.py --config [Name of configuration file]
For example, to reproduce the classification results on the Enron dataset using our default settings in Table 1, please execute the following command:
$ python RICC.py --config Enron_equal_close_ENM.yaml
$ python RICC.py --config Enron_equal_close_NNI.yaml
For more detailed examples, please refer to this link.
This research project has been conducted by WSP Lab at KAIST.
To cite our paper:
@INPROCEEDINGS{shin:www:2023,
author = {Dongwon Shin and Suyoung Lee and Sooel Son},
title = {{RICC}: Robust Collective Classification of Sybil Accounts},
booktitle = {Proceedings of the {ACM} Web Conference},
pages = {2329--2339},
year = 2023
}