How You Act Tells a Lot: Privacy-Leaking Attack on Deep Reinforcement Learning.
Xinlei PanWeiyao WangXiaoshuai ZhangBo LiJinfeng YiDawn SongPublished in: AAMAS (2019)
Keyphrases
- reinforcement learning
- sensitive information
- security risks
- privacy preserving
- information leakage
- authentication protocol
- malicious users
- personal information
- third party
- function approximation
- private information
- model free
- privacy preserving data mining
- secret key
- markov decision processes
- deep learning
- privacy preservation
- multi agent
- dynamic programming
- reinforcement learning algorithms
- sensitive data
- security measures
- privacy protection
- state space
- countermeasures
- differential privacy
- learning algorithm
- privacy concerns
- data privacy
- optimal policy
- private data
- optimal control
- machine learning
- privacy aware
- privacy issues
- function approximators
- neural network
- learning process
- personal data
- location privacy
- action selection
- legal issues
- security issues