Prior Convictions: Black-box Adversarial Attacks with Bandits and Priors.
Andrew IlyasLogan EngstromAleksander MadryPublished in: ICLR (Poster) (2019)
Keyphrases
- black box
- prior knowledge
- prior information
- black boxes
- white box
- prior probabilities
- prior distribution
- maximum a posteriori
- prior model
- countermeasures
- white box testing
- hybrid systems
- smoothness prior
- bayesian framework
- bayesian model
- watermarking scheme
- test cases
- rule extraction
- prior models
- integration testing
- dirichlet prior
- database
- state transition
- dirichlet process
- stochastic systems
- open source
- web services
- machine learning