safety security
Encrypted Reinforcement Learning Stopped Drifting
A zero-mean projection kept polynomial approximations inside their safe range without ciphertext bootstrapping.
Summary
A zero-mean projection kept polynomial approximations inside their safe range without ciphertext bootstrapping.
Fully homomorphic reinforcement learning replaces nonlinear operations with polynomials, which can recursively amplify Bellman error. The Homomorphic Advantage Operator centers temporal-difference targets to remove the uniform value baseline that drives this drift while preserving action rankings. Across tabular, encrypted CartPole and logistics tests, reported agents had zero boundary breaches in all seeds, versus breaches in three of five regularized seeds and 83.8 percent of unstabilized episodes. These controlled results do not establish production security or broad RL performance.
Why it matters
A zero-mean projection kept polynomial approximations inside their safe range without ciphertext bootstrapping.
Limits and context
- These controlled results do not establish production security or broad RL performance.
Key claims
A zero-mean projection kept polynomial approximations inside their safe range without ciphertext bootstrapping.
Qualification: These controlled results do not establish production security or broad RL performance.
Evidence: source-2026-10-02-016
Sources
- arXiv preprint 2610.02074arXiv · primary research
Corrections
No corrections have been recorded for this story.