Multi-level Certified Defense Against Poisoning Attacks in Offline Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Shijie, Cullen, Andrew C., Montague, Paul, Erfani, Sarah, Rubinstein, Benjamin I. P.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908380654731264
author Liu, Shijie
Cullen, Andrew C.
Montague, Paul
Erfani, Sarah
Rubinstein, Benjamin I. P.
author_facet Liu, Shijie
Cullen, Andrew C.
Montague, Paul
Erfani, Sarah
Rubinstein, Benjamin I. P.
contents Similar to other machine learning frameworks, Offline Reinforcement Learning (RL) is shown to be vulnerable to poisoning attacks, due to its reliance on externally sourced datasets, a vulnerability that is exacerbated by its sequential nature. To mitigate the risks posed by RL poisoning, we extend certified defenses to provide larger guarantees against adversarial manipulation, ensuring robustness for both per-state actions, and the overall expected cumulative reward. Our approach leverages properties of Differential Privacy, in a manner that allows this work to span both continuous and discrete spaces, as well as stochastic and deterministic environments -- significantly expanding the scope and applicability of achievable guarantees. Empirical evaluations demonstrate that our approach ensures the performance drops to no more than $50\%$ with up to $7\%$ of the training data poisoned, significantly improving over the $0.008\%$ in prior work~\citep{wu_copa_2022}, while producing certified radii that is $5$ times larger as well. This highlights the potential of our framework to enhance safety and reliability in offline RL.
format Preprint
id arxiv_https___arxiv_org_abs_2505_20621
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Multi-level Certified Defense Against Poisoning Attacks in Offline Reinforcement Learning
Liu, Shijie
Cullen, Andrew C.
Montague, Paul
Erfani, Sarah
Rubinstein, Benjamin I. P.
Machine Learning
Artificial Intelligence
Similar to other machine learning frameworks, Offline Reinforcement Learning (RL) is shown to be vulnerable to poisoning attacks, due to its reliance on externally sourced datasets, a vulnerability that is exacerbated by its sequential nature. To mitigate the risks posed by RL poisoning, we extend certified defenses to provide larger guarantees against adversarial manipulation, ensuring robustness for both per-state actions, and the overall expected cumulative reward. Our approach leverages properties of Differential Privacy, in a manner that allows this work to span both continuous and discrete spaces, as well as stochastic and deterministic environments -- significantly expanding the scope and applicability of achievable guarantees. Empirical evaluations demonstrate that our approach ensures the performance drops to no more than $50\%$ with up to $7\%$ of the training data poisoned, significantly improving over the $0.008\%$ in prior work~\citep{wu_copa_2022}, while producing certified radii that is $5$ times larger as well. This highlights the potential of our framework to enhance safety and reliability in offline RL.
title Multi-level Certified Defense Against Poisoning Attacks in Offline Reinforcement Learning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.20621