Gradient-Based Clean Label Backdoor Attack to Graph Neural Networks.
Ryo MeguroHiroya KatoShintaro NarisadaSeira HidanoKazuhide FukushimaTakuo SuganumaMasahiro HijiPublished in: ICISSP (2024)
Keyphrases
- neural network
- label propagation
- attack graph
- graph model
- pattern recognition
- graph representation
- countermeasures
- graph structure
- graph theory
- structured data
- graph mining
- genetic algorithm
- graph matching
- directed graph
- multi label
- random walk
- fuzzy logic
- class labels
- feed forward
- back propagation
- training process
- directed acyclic graph
- semi supervised
- training data
- self organizing maps
- graph based algorithm
- bayesian networks
- graph data
- artificial neural networks
- activation function
- spanning tree
- multi layer
- weighted graph
- labeled data
- semi supervised learning
- fault diagnosis