Adversarial Attack Protection Scalar Multiplication for WSNs Resistance Machine-Learning Side-channel Attack

Abstract

Wireless sensor networks (WSNs) have limited storage and low processing capabilities. However, these devices may be captured by opponents, so the security requirements are particularly strict. With the development of side-channel attacks based on the machine-learning, designing a safe elliptic curve scalar multiplication for computing-limited devices like WSNs has become a major challenge. Based on the adversarial attack technology, a secure scalar multiplication is proposed in this article. The main contributions are: (1) We propose an efficient non-zero form (NZF) encoding algorithm that can be applied to various types of elliptic curves; (2) we have designed a secure scalar multiplication algorithm that can resist against conventional side-channel attacks such as SPA, DA, DPA, RPA, and ZPA; and (3) we propose an adversarial protection mechanism based on blind point technology and NZF coding, which can prevent side-channel attacks based on machine learning. The algorithm has no precomputation and is suitable for low communication frequency, low calculation amount, and high security requirements. Especially, it can be applied to lightweight equipment such as WSN and IoT.