Core Concepts
Incorporating bootstrapping into encrypted control systems ensures stability and performance through robust control analysis.
Abstract
The content discusses the challenges of encrypted dynamic controllers, focusing on bootstrapping to avoid errors and scaling factors. It provides insights into the analysis of dynamic encrypted control, the use of homomorphic encryption, and the impact of bootstrapping errors on system stability and performance. The paper introduces a novel approach to incorporate bootstrapping into the analysis, reducing conservatism and enhancing control performance.
I. Introduction
Encrypted control for outsourcing computations.
Homomorphic encryption for secure computations.
II. Preliminaries
Notation and properties of cryptosystems.
Introduction to CKKS scheme for approximate real numbers.
III. Bootstrapping Polynomial
Error approximation for bootstrapping.
Relative error measure for stability in encrypted control.
IV. Problem Formulation
System description and robust control framework.
Stability and performance analysis with bootstrapping.
V. Dynamic Control with Bootstrapping
Analysis of lifted dynamics for less conservative testing.
Numerical example and empirical evaluation.
VI. Analysis of Reset and FIR Controllers
Application of Theorem 2 to reset and FIR controllers.
Stability and performance tests for dynamic controllers.
VII. Summary and Outlook
Incorporating bootstrapping in encrypted control analysis.
Benefits of tailored bootstrapping polynomials for control performance.
Stats
"The proof follows directly from our uncertainty description for bootstrapping (13), which suits the robust stability and performance test in [20, Thm. 10.4.]."
"The proof follows from Lemma 1 and Theorem 1."
"Using the polynomial from Section III with TBS = 10, an upper bound on the ℓ2-gain is found with Qp = −γ2 ℓ2I,Sp = 0, and Rp = I."
Quotes
"This is the first paper to explicitly incorporate the bootstrapping effects into system analysis."
"Our analysis shows that encrypted control has different requirements on the bootstrapping than general purpose homomorphic encryption."