Bkd-FedGNN

Bkd-FedGNN is a benchmark designed to evaluate classification backdoor attacks on Federated Graph Neural Networks by decomposing the attack process into trigger generation and injection steps, and addressing both node-level and graph-level classification tasks.

Papers using Bkd-FedGNN (1)

Bkd-FedGNN: A Benchmark for Classification Backdoor Attacks on Federated Graph Neural Network2023