← 返回 JSSC 论文列表JSSC 2023第1期Digital Circuits28nmNeural Network Accelerator
A Threshold Implementation-Based Neural Network Accelerator With Power and Elect
基于阈值实现的神经网络加速器,有效抵御功耗和电磁侧信道攻击。
28nm CMOS, 0.95V, 125MHz
神经网络加速器侧信道安全阈值实现功耗分析电磁攻击
▸创新点1:阈值实现(TI)掩码技术的电路创新,通过将敏感数据分割为多个共享部分并在不同时钟周期处理,有效抵御高阶侧信道攻击,实测安全性能超过200万次攻击测试。
▸创新点2:模型参数加密的系统创新,采用轻量级加密算法保护存储中的神经网络权重,结合TI掩码实现端到端安全,面积开销仅增加64%。
▸创新点3:针对水平功耗分析(HPA)攻击的防护方法创新,设计时序随机化数据调度单元,打破功耗与操作数据的相关性,能量开销控制在5.5倍以内。
▸创新点4:28nm工艺下实现125MHz/0.95V的高能效架构,首次在神经网络加速器中整合TI掩码、加密和HPA防护,达到0.159mm²的紧凑面积。
Abstract
With the recent advancements in machine learning
(ML) theory, a lot of energy-efficient neural network (NN)
accelerators have been developed. However, their associated side-
channel security vulnerabilities pose a major concern. There
have been several proof-of-concept attacks demonstrating the
extraction of their model parameters and input data. This work
introduces a threshold implementation (TI) masking-based NN
accelerator that secures mod el parameters and inputs against
power and electromag