Skip to content

Latest commit

 

History

History
80 lines (66 loc) · 2.89 KB

README.md

File metadata and controls

80 lines (66 loc) · 2.89 KB
Raw

Configuration

Configuration Form

A configuration file is a yaml file with the following fields. We also provide the configuration files used in our evaluation.

  • Configuration fields for RICC

    • case: Identifier for each experiment. You must use a different name for each experiment.
    • graph: Name of the graph dataset. (Enron, Facebook, Twitter_small, Twitter_large).
    • edge_manipulation_cost: Type of the edge manipulation cost. (equal, uni, cat).
    • target_node_type: Type of the target node selection method. (close, rand, lcc).
    • attack: Type of the attack. (ENM, NNI).
    • epoch: Maximum number of the epoch. We use the term iteration instead of epoch in our paper.
    • lr: The learning rate. We use the term $\epsilon$ in our paper.
    • buffer: The buffer ratio. We use the term $\rho$ in our paper.
    • sampling_size: The number of sampled nodes. We use the term $N$ in our paper.
    • interval: The epoch interval for printing and recording metrics during the experiment.
    • num_modified_edges: The number of modified edges.
    • num_added_nodes: The number of added edges.
    • num_target_nodes: The number of target nodes.

  • Configuration fields for collective classification

    • theta: The absolute value used for prior scores. We use the term $\theta$ in our paper.
    • weight: The weight value used for collective classification.
    • iteration: The maximum number of iterations used for collective classification. We use the term $T_{max}$ in our paper.
    • threshold: The threshold value used to classify Sybil nodes.

Examples

To reproduce the classification result on the Enron dataset using our default setting in Table 1, please run:

$ python RICC.py --config Enron_equal_close_ENM.yaml

$ python RICC.py --config Enron_equal_close_NNI.yaml

To reproduce the classification result on the Enron graph manipulated using different attack budgets, as shown in Figures 3 & 4, please run:

$ python RICC.py --config Enron_equal_close_ENM_K_80.yaml

$ python RICC.py --config Enron_equal_close_NNI_N_110.yaml

To reproduce the classification result on the Enron graph manipulated using a different number of target nodes, as shown in Figure 6, please run:

$ python RICC.py --config Enron_equal_close_ENM_T_600.yaml

To reproduce the classification result on the Enron graph manipulated using different attack strategies, as shown in Tables 2 & 3, please run:

$ python RICC.py --config Enron_equal_lcc_ENM.yaml

$ python RICC.py --config Enron_equal_rand_ENM.yaml

$ python RICC.py --config Enron_cat_close_ENM.yaml

$ python RICC.py --config Enron_uni_close_ENM.yaml

If you want to run RICC with the Facebook dataset, please run:

$ python RICC.py --config Facebook_equal_close_ENM.yaml

$ python RICC.py --config Facebook_equal_close_NNI.yaml