Defense Against Reward Poisoning Attacks in Reinforcement Learning.
Kiarash BanihashemAdish SinglaGoran RadanovicPublished in: Trans. Mach. Learn. Res. (2023)
Keyphrases
- reinforcement learning
- defense mechanisms
- ddos attacks
- function approximation
- countermeasures
- state space
- eligibility traces
- intrusion detection
- computer virus
- markov decision processes
- network security
- learning algorithm
- reinforcement learning algorithms
- watermarking algorithm
- model free
- action selection
- multi agent
- reward function
- dynamic programming
- computer security
- malicious attacks
- optimal policy
- security threats
- machine learning
- watermarking scheme
- partially observable environments
- total reward
- security risks
- information security
- dos attacks
- average reward
- cyber attacks
- reinforcement learning methods
- action space
- policy gradient
- policy evaluation
- policy iteration
- denial of service attacks
- markov decision process
- watermarking technique
- partially observable
- temporal difference
- long run