一句话：聚焦从硬件体系结构事件中提取 DNN 模型架构，并指出现有方法在先验知识、鲁棒性与通用性上的不足。
随着深度神经网络应用范围扩大，DNN 模型架构因知识产权保护或对抗攻击风险而变得敏感。已有研究尝试利用硬件平台中的体系结构级事件提取模型架构信息，但存在以下限制：需要被攻击模型的先验知识、鲁棒性和通用性不足，或只能获得被攻击模型的不完整架构信息。本文以 DeepSniffer 为名，针对上述问题展开研究，目标是在减少先验依赖的同时，从硬件体系结构事件中提取 DNN 模型架构。给出的摘要未描述具体方法、实验设置与量化结果，因此无法进一步概括其技术细节和评估结论。
As deep neural networks (DNNs) continue their reach into a wide range of application domains, the neural network architecture of DNN models becomes an increasingly sensitive subject, due to either intellectual property protection or risks of adversarial attacks. Previous studies explore to leverage architecture-level events disposed in hardware platforms to extract the model architecture information. They pose the following limitations: requiring a priori knowledge of victim models, lacking in robustness and generality, or obtaining incomplete information of the victim model architecture.
@inproceedings{paperbot5442,
title = {DeepSniffer},
author = {Xing Hu and Ling Liang and Shuangchen Li and Lei Deng and Pengfei Zuo and Yu Ji and Xinfeng Xie and Yufei Ding and Chang Liu and Timothy Sherwood and Yuan Xie},
booktitle = {ASPLOS},
year = {2020},
doi = {10.1145/3373376.3378460}
}