HideNSeek

Source code of KDD'25 submission "On Measuring Unnoticeability of Graph Adversarial Attacks: Observations, New Measure, and Applications"

Environments

• python 3.7.11
• numpy==1.21.2
• torch==1.10.0 (with CUDA 11.3)
• sklearn==1.0.2
• scipy==1.7.3

Generate Attack Graph

example: cora dataset / attack method: PGD attack / attack rate: 10%

python attack.py --graph_name cora --attacker pgd --attack_rate 10 --seed 0

python attack.py --graph_name cora --attacker pgd --attack_rate 10 --seed 1

python attack.py --graph_name cora --attacker pgd --attack_rate 10 --seed 2

python attack.py --graph_name cora --attacker pgd --attack_rate 10 --seed 3

python attack.py --graph_name cora --attacker pgd --attack_rate 10 --seed 4

Measure Noticeability Score by HideNSeek (Q1)

example: cora dataset / attack method: PGD attack / attack rate: 10%

python q1_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 0

Noticeability score curve according to the node classification accuracy (Q2)

example: cora dataset / attack method: PGD attack / attack rate: 10% / average over seed 0 ~ 4

python q2_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 0

python q2_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 1

python q2_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 2

python q2_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 3

python q2_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 4

python q2_plot.py --graph_name cora --attacker pgd --attack_rate 10

Further Application (Q3)

example: cora dataset / attack method: PGD attack / attack rate: 10% / GCN type: MedianGCN

python q3_main.py --graph_name cora --attacker pgd --attack_rate 10 --seed 0 --gcn_name mediangcn