Deceptive Kernel Function on Observations of Discrete POMDP
This addresses security threats to AI agents in partially observable environments, though it appears incremental as it applies deception to existing POMDP frameworks.
The paper introduces a deceptive kernel function that falsifies observations in discrete POMDPs, showing it misleads agent beliefs and substantially lowers rewards in experiments on two POMDP problems.
This paper studies the deception applied on agent in a partially observable Markov decision process. We introduce deceptive kernel function (the kernel) applied to agent's observations in a discrete POMDP. Based on value iteration, value function approximation and POMCP three characteristic algorithms used by agent, we analyze its belief being misled by falsified observations as the kernel's outputs and anticipate its probable threat on agent's reward and potentially other performance. We validate our expectation and explore more detrimental effects of the deception by experimenting on two POMDP problems. The result shows that the kernel applied on agent's observation can affect its belief and substantially lower its resulting rewards; meantime certain implementation of the kernel could induce other abnormal behaviors by the agent.