Encrypting Controller using Fully Homomorphic Encryption for Security of Cyber-Physical Systems**The work of J. Kim, C. Lee, and H. Shim was supported by ICT R & D program of MSIP/IITP Grant number 14-824-09-013, Resilient Cyber-Physical Systems Research. The work of J. H. Cheon, A. Kim, M. Kim, and Y. Song was supported by IT R & D program of MSIP/KEIT [No. 0450-21060006] and Samsung Electronics Co., Ltd. (No. 0421-20150074).

TLDR

Protecting sensor‑to‑controller and controller‑to‑actuator signals is critical for cyber‑physical system security, yet encryption requires decryption for computation, exposing the controller’s secret key. The study proposes using fully homomorphic encryption to enable encrypted computation in cyber‑physical controllers without decryption. The authors implement this by running multiple controllers orchestrated together and employing a tree‑based sequential matrix multiplication to extend encrypted variable lifespan. The approach is validated on a quadruple water tank system, demonstrating its effectiveness.

Abstract

In order to enhance security of cyber-physical systems, it is important to protect the signals from sensors to the controller, and from the controller to the actuator, because the attackers often steal and compromise those signals. One immediate solution could be encrypting the signals, but in order to perform computation in the controller, they should be decrypted before computation and encrypted again after computation. For this, the controller keeps the secret key, which in turn increases vulnerability from the attacker. In this paper, we introduce the fully homomorphic encryption (FHE), which is an advanced cryptography that has enabled arithmetic operations directly on the encrypted variables without decryption. However, this also introduces several new issues that have not been studied for conventional controllers. Most of all, an encrypted variable has a finite lifespan, which decreases as an arithmetic operation is performed on it. Our solution is to run multiple controllers, and orchestrate them systematically. Also, in order to slow down the decrease of the lifespan, a tree-based computation of sequential matrix multiplication is introduced. We finally demonstrate the effectiveness of the proposed algorithm with quadruple water tank example.