Adversarial observations in probabilistic State-Space Models for robust Reinforcement Learning
M. Santos-Pascual, D. Ríos Insua
stat.ML, cs.LG, stat.ME
Submitted: 2026-06-18
Comments: 42 pages, 10 figures, PREPRINT ONGOING: Revised version with additional theoretical results and experiments
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
The gist: Decision-making under partial or adversarial observability requires accurate inference of the environment's latent state and its associated uncertainty.
Terminology
Abstract
Decision-making under partial or adversarial observability requires accurate inference of the environment's latent state and its associated uncertainty. This work analyses adversarial attacks on linear state-space models, where the attacker alters observations subject to likelihood constraints that ensure that the perturbations remain statistically consistent with the observation model. We analyse how such adversarial yet plausible observations shift inference about latent states and affect downstream decision-making and the performance of reinforcement learning agents. In addition, we introduce an online Bayesian defence based on directional covariance adaptation, which selectively reduces the influence of observations by comparing their estimated impact with that of the computed most disruptive direction, while preserving information in the remaining orthogonal observation subspace. The proposed framework provides a principled approach to constructing robust inference and decision-making systems, with direct relevance to safety-critical applications such as robotics, where reliable operation under sensor noise, partial failures, and adversarial conditions is essential.
Sources
- Explaining and Harnessing Adversarial Examples
- Adversarial Attacks on Neural Network Policies
- Distributionally Robust Kalman Filtering over Finite and Infinite Horizon
- Towards Deep Learning Models Resistant to Adversarial Attacks
- Proximal Policy Optimization Algorithms
- Advancing Intelligent Sequence Modeling: Evolution, Trade-offs, and Applications of State-Space Architectures from S4 to Mamba
Related papers
- Behavior of prediction performance metrics with rare events
- Optimal Estimation of Generic Dynamics by Path-Dependent Neural Jump ODEs
- A Posterior-Dynamics Framework for Imaging Inverse Problems with Pretrained Diffusion Priors
- Scalable Posterior Uncertainty for Flexible Density-Based Clustering
- A Transport Geometry of Benign Overfitting
- Efficient quadratic entropy with distance sketches