Gotta Catch 'Em All: Using Concealed Trapdoors to Detect Adversarial Attacks on Neural Networks.
Shawn ShanEmily WillsonBolun WangBo LiHaitao ZhengBen Y. ZhaoPublished in: CoRR (2019)
Keyphrases
- neural network
- detection algorithm
- expectation maximization
- network attacks
- automatic detection
- countermeasures
- pattern recognition
- multi agent
- image segmentation
- artificial neural networks
- detecting malicious
- fuzzy logic
- maximum likelihood
- em algorithm
- unsupervised learning
- detection method
- watermarking algorithm
- neural network model
- denial of service attacks
- security threats
- intrusion detection system
- security protocols
- detect malicious
- malicious attacks
- training process
- watermarking scheme
- recurrent neural networks
- feed forward
- fault diagnosis
- back propagation
- learning algorithm
- genetic algorithm