Skip to content

Front edge prediction stage3 #104

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
LukyanovKirillML merged 2 commits into from
Feb 6, 2026
Merged

Front edge prediction stage3 #104

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5b01aa4
training edge pred ok
mishadr Feb 4, 2026
09d7204
fix frontend choose edge
mishadr Feb 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
122 changes: 61 additions & 61 deletions experiments/attack_defense_exps.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -169,7 +169,7 @@ def test_attack_defense():
_import_path=EVASION_ATTACK_PARAMETERS_PATH,
_config_class="EvasionAttackConfig",
_config_kwargs={
"node_idx": 0, # Node for attack
"element_idx": 0, # Node for attack
"n_perturbations": 20,
"perturb_features": True,
"perturb_structure": True,
Expand Down Expand Up @@ -415,10 +415,10 @@ def test_nettack_evasion():
save_model_flag=False)

# Node for attack
node_idx = 1
element_idx = 1

# Evaluate model
mask_loc = Metric.create_mask_by_target_list(y_true=dataset.labels, target_list=[node_idx])
mask_loc = Metric.create_mask_by_target_list(y_true=dataset.labels, target_list=[element_idx])
acc_test_loc = gnn_model_manager.evaluate_model(gen_dataset=dataset,
metrics=[Metric("Accuracy", mask=mask_loc)])[mask_loc]['Accuracy']

Expand All @@ -435,11 +435,11 @@ def test_nettack_evasion():
with torch.no_grad():
probabilities = torch.exp(gnn_model_manager.gnn(dataset.data.x, dataset.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_before_evasion_attack = {"node_idx": node_idx,
info_before_evasion_attack = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand All @@ -450,7 +450,7 @@ def test_nettack_evasion():
_import_path=EVASION_ATTACK_PARAMETERS_PATH,
_config_class="EvasionAttackConfig",
_config_kwargs={
"node_idx": node_idx,
"element_idx": element_idx,
"n_perturbations": 20,
"perturb_features": True,
"perturb_structure": True,
Expand All @@ -469,11 +469,11 @@ def test_nettack_evasion():
probabilities = torch.exp(gnn_model_manager.gnn(gnn_model_manager.evasion_attacker.attack_diff.data.x,
gnn_model_manager.evasion_attacker.attack_diff.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_after_evasion_attack = {"node_idx": node_idx,
info_after_evasion_attack = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand Down Expand Up @@ -545,18 +545,18 @@ def test_qattack():
print(f"Accuracy on test: {acc_test}")

# Node for attack
# node_idx = 0
# element_idx = 0
#
# # Model prediction on a node before an evasion attack on it
# gnn_model_manager.gnn.eval()
# with torch.no_grad():
# probabilities = torch.exp(gnn_model_manager.gnn(dataset.data.x, dataset.data.edge_index))
#
# predicted_class = probabilities[node_idx].argmax().item()
# predicted_probability = probabilities[node_idx][predicted_class].item()
# real_class = dataset.data.y[node_idx].item()
# predicted_class = probabilities[element_idx].argmax().item()
# predicted_probability = probabilities[element_idx][predicted_class].item()
# real_class = dataset.data.y[element_idx].item()

# info_before_evasion_attack = {"node_idx": node_idx,
# info_before_evasion_attack = {"element_idx": element_idx,
# "predicted_class": predicted_class,
# "predicted_probability": predicted_probability,
# "real_class": real_class}
Expand All @@ -577,11 +577,11 @@ def test_qattack():
# probabilities = torch.exp(gnn_model_manager.gnn(gnn_model_manager.evasion_attacker.attack_diff.data.x,
# gnn_model_manager.evasion_attacker.attack_diff.data.edge_index))
#
# predicted_class = probabilities[node_idx].argmax().item()
# predicted_probability = probabilities[node_idx][predicted_class].item()
# real_class = dataset.data.y[node_idx].item()
# predicted_class = probabilities[element_idx].argmax().item()
# predicted_probability = probabilities[element_idx][predicted_class].item()
# real_class = dataset.data.y[element_idx].item()
#
# info_after_evasion_attack = {"node_idx": node_idx,
# info_after_evasion_attack = {"element_idx": element_idx,
# "predicted_class": predicted_class,
# "predicted_probability": predicted_probability,
# "real_class": real_class}
Expand Down Expand Up @@ -834,18 +834,18 @@ def test_pgd():
print(f"Accuracy on test: {acc_test}")

# Node for attack
node_idx = 650
element_idx = 650

# Model prediction on a node before PGD attack on it
gnn_model_manager.gnn.eval()
with torch.no_grad():
probabilities = torch.exp(gnn_model_manager.gnn(dataset.data.x, dataset.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_before_pgd_attack_on_node = {"node_idx": node_idx,
info_before_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand All @@ -857,7 +857,7 @@ def test_pgd():
_config_class="EvasionAttackConfig",
_config_kwargs={
"is_feature_attack": True,
"element_idx": node_idx,
"element_idx": element_idx,
"epsilon": 0.1,
"learning_rate": 0.001,
"num_iterations": 500,
Expand All @@ -876,11 +876,11 @@ def test_pgd():
probabilities = torch.exp(gnn_model_manager.gnn(gnn_model_manager.evasion_attacker.attack_diff.data.x,
gnn_model_manager.evasion_attacker.attack_diff.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_after_pgd_attack_on_node = {"node_idx": node_idx,
info_after_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand Down Expand Up @@ -1029,18 +1029,18 @@ def test_pgd_structure():
print(f"Accuracy on test: {acc_test}")

# Node for attack
node_idx = 1700
element_idx = 1700

# Model prediction on a node before PGD attack on it
gnn_model_manager.gnn.eval()
with torch.no_grad():
probabilities = torch.exp(gnn_model_manager.gnn(dataset.data.x, dataset.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_before_pgd_attack_on_node = {"node_idx": node_idx,
info_before_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand All @@ -1052,7 +1052,7 @@ def test_pgd_structure():
_config_class="EvasionAttackConfig",
_config_kwargs={
"is_feature_attack": False,
"element_idx": node_idx,
"element_idx": element_idx,
"num_iterations": 30,
}
)
Expand All @@ -1068,11 +1068,11 @@ def test_pgd_structure():
probabilities = torch.exp(gnn_model_manager.gnn(gnn_model_manager.evasion_attacker.attack_diff.data.x,
gnn_model_manager.evasion_attacker.attack_diff.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_after_pgd_attack_on_node = {"node_idx": node_idx,
info_after_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand Down Expand Up @@ -1219,18 +1219,18 @@ def test_fgsm():
print(f"Accuracy on test: {acc_test}")

# Node for attack
node_idx = 650
element_idx = 650

# Model prediction on a node before PGD attack on it
gnn_model_manager.gnn.eval()
with torch.no_grad():
probabilities = torch.exp(gnn_model_manager.gnn(dataset.data.x, dataset.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_before_pgd_attack_on_node = {"node_idx": node_idx,
info_before_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand All @@ -1242,7 +1242,7 @@ def test_fgsm():
_config_class="EvasionAttackConfig",
_config_kwargs={
"is_feature_attack": False,
"element_idx": node_idx,
"element_idx": element_idx,
"epsilon": 0.5,
}
)
Expand All @@ -1258,11 +1258,11 @@ def test_fgsm():
probabilities = torch.exp(gnn_model_manager.gnn(gnn_model_manager.evasion_attacker.attack_diff.data.x,
gnn_model_manager.evasion_attacker.attack_diff.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_after_pgd_attack_on_node = {"node_idx": node_idx,
info_after_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand Down Expand Up @@ -1411,18 +1411,18 @@ def test_rewatt():
print(f"Accuracy on test: {acc_test}")

# Node for attack
node_idx = 44
element_idx = 44

# Model prediction on a node before ReWatt attack on it
gnn_model_manager.gnn.eval()
with torch.no_grad():
probabilities = torch.exp(gnn_model_manager.gnn(dataset.data.x, dataset.data.edge_index))

predicted_class = probabilities[node_idx].argmax().item()
predicted_probability = probabilities[node_idx][predicted_class].item()
real_class = dataset.data.y[node_idx].item()
predicted_class = probabilities[element_idx].argmax().item()
predicted_probability = probabilities[element_idx][predicted_class].item()
real_class = dataset.data.y[element_idx].item()

info_before_pgd_attack_on_node = {"node_idx": node_idx,
info_before_pgd_attack_on_node = {"element_idx": element_idx,
"predicted_class": predicted_class,
"predicted_probability": predicted_probability,
"real_class": real_class}
Expand All @@ -1433,7 +1433,7 @@ def test_rewatt():
_import_path=EVASION_ATTACK_PARAMETERS_PATH,
_config_class="EvasionAttackConfig",
_config_kwargs={
"element_idx": node_idx,
"element_idx": element_idx,
"eps": 0.001,
"epochs": 10,
}
Expand All @@ -1450,11 +1450,11 @@ def test_rewatt():
# probabilities = torch.exp(gnn_model_manager.gnn(gnn_model_manager.evasion_attacker.attack_diff.data.x,
# gnn_model_manager.evasion_attacker.attack_diff.data.edge_index))
#
# predicted_class = probabilities[node_idx].argmax().item()
# predicted_probability = probabilities[node_idx][predicted_class].item()
# real_class = dataset.data.y[node_idx].item()
# predicted_class = probabilities[element_idx].argmax().item()
# predicted_probability = probabilities[element_idx][predicted_class].item()
# real_class = dataset.data.y[element_idx].item()
#
# info_after_pgd_attack_on_node = {"node_idx": node_idx,
# info_after_pgd_attack_on_node = {"element_idx": element_idx,
# "predicted_class": predicted_class,
# "predicted_probability": predicted_probability,
# "real_class": real_class}
Expand Down
Loading