Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035388736> ?p ?o ?g. }
Showing items 1 to 73 of
73
with 100 items per page.
- W3035388736 endingPage "7984" @default.
- W3035388736 startingPage "7974" @default.
- W3035388736 abstract "We study a security threat to reinforcement learning where an attacker poisons the learning environment to force the agent into executing a target policy chosen by the attacker. As a victim, we consider RL agents whose objective is to find a policy that maximizes average reward in undiscounted infinite-horizon problem settings. The attacker can manipulate the rewards or the transition dynamics in the learning environment at training-time and is interested in doing so in a stealthy manner. We propose an optimization framework for finding an emph{optimal stealthy attack} for different measures of attack cost. We provide sufficient technical conditions under which the attack is feasible and provide lower/upper bounds on the attack cost. We instantiate our attacks in two settings: (i) an emph{offline} setting where the agent is doing planning in the poisoned environment, and (ii) an emph{online} setting where the agent is learning a policy using a regret-minimization framework with poisoned feedback. Our results show that the attacker can easily succeed in teaching any target policy to the victim under mild conditions and highlight a significant security threat to reinforcement learning agents in practice." @default.
- W3035388736 created "2020-06-19" @default.
- W3035388736 creator A5005193721 @default.
- W3035388736 creator A5027711113 @default.
- W3035388736 creator A5047197460 @default.
- W3035388736 creator A5052833351 @default.
- W3035388736 creator A5073837056 @default.
- W3035388736 date "2020-07-12" @default.
- W3035388736 modified "2023-09-24" @default.
- W3035388736 title "Policy Teaching via Environment Poisoning: Training-time Adversarial Attacks against Reinforcement Learning" @default.
- W3035388736 hasPublicationYear "2020" @default.
- W3035388736 type Work @default.
- W3035388736 sameAs 3035388736 @default.
- W3035388736 citedByCount "15" @default.
- W3035388736 countsByYear W30353887362020 @default.
- W3035388736 countsByYear W30353887362021 @default.
- W3035388736 countsByYear W30353887362022 @default.
- W3035388736 crossrefType "proceedings-article" @default.
- W3035388736 hasAuthorship W3035388736A5005193721 @default.
- W3035388736 hasAuthorship W3035388736A5027711113 @default.
- W3035388736 hasAuthorship W3035388736A5047197460 @default.
- W3035388736 hasAuthorship W3035388736A5052833351 @default.
- W3035388736 hasAuthorship W3035388736A5073837056 @default.
- W3035388736 hasConcept C119857082 @default.
- W3035388736 hasConcept C126255220 @default.
- W3035388736 hasConcept C154908896 @default.
- W3035388736 hasConcept C154945302 @default.
- W3035388736 hasConcept C28761237 @default.
- W3035388736 hasConcept C33923547 @default.
- W3035388736 hasConcept C37736160 @default.
- W3035388736 hasConcept C38652104 @default.
- W3035388736 hasConcept C41008148 @default.
- W3035388736 hasConcept C50817715 @default.
- W3035388736 hasConcept C97541855 @default.
- W3035388736 hasConceptScore W3035388736C119857082 @default.
- W3035388736 hasConceptScore W3035388736C126255220 @default.
- W3035388736 hasConceptScore W3035388736C154908896 @default.
- W3035388736 hasConceptScore W3035388736C154945302 @default.
- W3035388736 hasConceptScore W3035388736C28761237 @default.
- W3035388736 hasConceptScore W3035388736C33923547 @default.
- W3035388736 hasConceptScore W3035388736C37736160 @default.
- W3035388736 hasConceptScore W3035388736C38652104 @default.
- W3035388736 hasConceptScore W3035388736C41008148 @default.
- W3035388736 hasConceptScore W3035388736C50817715 @default.
- W3035388736 hasConceptScore W3035388736C97541855 @default.
- W3035388736 hasOpenAccess W3035388736 @default.
- W3035388736 hasRelatedWork W1931078822 @default.
- W3035388736 hasRelatedWork W2020764470 @default.
- W3035388736 hasRelatedWork W2182055801 @default.
- W3035388736 hasRelatedWork W2786676179 @default.
- W3035388736 hasRelatedWork W2890752237 @default.
- W3035388736 hasRelatedWork W2917822786 @default.
- W3035388736 hasRelatedWork W2964725706 @default.
- W3035388736 hasRelatedWork W2966120739 @default.
- W3035388736 hasRelatedWork W2970023415 @default.
- W3035388736 hasRelatedWork W2970912396 @default.
- W3035388736 hasRelatedWork W2981396729 @default.
- W3035388736 hasRelatedWork W3009289457 @default.
- W3035388736 hasRelatedWork W3011708929 @default.
- W3035388736 hasRelatedWork W3013223143 @default.
- W3035388736 hasRelatedWork W3034593529 @default.
- W3035388736 hasRelatedWork W3049367159 @default.
- W3035388736 hasRelatedWork W3094116649 @default.
- W3035388736 hasRelatedWork W3109594330 @default.
- W3035388736 hasRelatedWork W3209580873 @default.
- W3035388736 hasRelatedWork W3127643357 @default.
- W3035388736 hasVolume "1" @default.
- W3035388736 isParatext "false" @default.
- W3035388736 isRetracted "false" @default.
- W3035388736 magId "3035388736" @default.
- W3035388736 workType "article" @default.