Certifying Safety in Reinforcement Learning under Adversarial Perturbation Attacks.
Junlin WuHussein SibaiYevgeniy VorobeychikPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- multi agent
- function approximation
- reinforcement learning algorithms
- countermeasures
- model free
- markov decision processes
- learning algorithm
- malicious attacks
- state space
- supervised learning
- traffic analysis
- machine learning
- malicious users
- temporal difference learning
- temporal difference
- optimal policy
- perturbation method
- learning classifier systems
- learning problems
- security protocols
- function approximators
- reinforcement learning methods
- security threats
- learning process
- denial of service attacks
- multi agent reinforcement learning
- multi agent systems
- attack detection
- neural network
- chosen plaintext