计算机科学
量子网络
量子密码学
量子
量子计算机
量子信息
安全传输
计算机网络
密码学
传输(电信)
加密
拓扑(电路)
量子信道
理论计算机科学
量子态
分布式计算
量子密钥分配
密码原语
计算
量子容量
量子信息科学
服务器
协议(科学)
量子算法
带着错误学习
编码(内存)
方案(数学)
构造(python库)
密码协议
掷硬币
国家(计算机科学)
作者
Wei‐Min Shi,Jia-Wei Yuan,Zi-Xuan Xu,Yu-Guang Yang,Yi‐Hua Zhou
出处
期刊:International Journal of Modern Physics C
[World Scientific]
日期:2026-11-19
卷期号:: 2750175-
标识
DOI:10.1142/s0129183127501750
摘要
Current approaches to secure gradient aggregation in quantum federated learning typically employ classical or quantum cryptographic techniques to encrypt local gradients prior to transmission to the server via classical or quantum channels. In this process, local gradient information must be transmitted as communication content to take part in aggregation, which increases the potential attack surface for privacy. To address this problem, this paper proposes a quantum secure gradient aggregation scheme based on Ancilla-Driven Quantum Computation (ADQC). The scheme eliminates the direct transmission of local gradient information via either classical or quantum channels. Clients remotely construct single-qubit rotation gates through ADQC, thereby encoding their local gradients into the quantum state held by the server and achieving secure aggregation of local gradients. The proposed scheme includes two protocols for secure gradient aggregation: A serial single-qubit protocol and a parallel multi-qubit protocol. Security analysis demonstrates that both protocols are resilient to external attacks and to attacks originating from servers and clients, thereby achieving privacy protection for local gradients.
科研通智能强力驱动
Strongly Powered by AbleSci AI