In this paper, a new cryptocurrency security protocol is proposed, which combines Galois group action and deep neural network (DNN), in order to improve the robustness of elliptic curve cryptography (ECC) in the face of mathematical attacks and side channel attacks. Traditional ECC protocols mostly rely on static parameters and are vulnerable to the weakness of curve structure, physical leakage and the threat of quantum computing. In this paper, firstly, the dynamic selection mechanism of ECC curve parameters is constructed by using Galois group action theory to ensure its high-order symmetry and anti-attack under group action; Secondly, a lightweight DNN model is designed to realize the intelligent identification of channel characteristics such as power consumption trajectory and improve the real-time detection ability of attack behavior. The experimental results show that the curve based on Galois constraint successfully resists all known mathematical attacks in 1000 tests, and the safety rate is improved from 93% of the traditional method to 100%. The DNN model achieves 99.1% F1-Score in the detection of side channel attacks, and the inference delay is as low as 2.8 ms. The overall security rate of the integrated system is improved to 99.2%, and the signature delay is only increased by 3ms, giving consideration to high security and practicality. This study provides a double-layer security framework of "mathematical defense+intelligent perception" for cryptocurrency security in the post-quantum era, which has strong theoretical value and application prospects.
Yixi Pan (Sun,) studied this question.