LinuxBoot固件深度学习框架移植
Deep learning framework migration scheme in LinuxBoot firmware environment
李玲 1赫俊民 1李珮玄 2孟魁2
作者信息
- 1. 中国石化股份有限公司胜利油田分公司物探研究院,山东 东营 257022
- 2. 上海交通大学网络空间安全学院,上海 200241
- 折叠
摘要
为使依赖深度学习框架的应用在LinuxBoot固件环境下发挥应有效能,提出一种在LinuxBoot固件环境中植入深度学习框架的技术方案.通过对Linux操作系统环境中正常运行深度学习框架所需依赖进行分析,设计一种将深度学习框架移植入LinuxBoot固件环境的技术路线,构建将二者合二为一的系统结构及交互流程,提出借助USB闪存驱动器存储移植内容的技术方案,给出具体的移植操作指令及初始化脚本.在Thinkpad固件中进行真机实验,展示移植后的深度学习框架在图像分类、自然语言处理、恶意代码检测、恶意代码分类等多种任务上均能正常运行,在各项评估指标上表现良好.
Abstract
To make applications that rely on deep learning framework run in the LinuxBoot firmware environment,a technical scheme for implanting deep learning framework in the LinuxBoot firmware environment was proposed.By analyzing the depen-dencies required for the normal operation of the deep learning framework in the Linux operating system environment,a technical route for porting the deep learning framework to the LinuxBoot firmware environment was designed.The system structure and interaction process that combined the deep learning framework and the LinuxBoot firmware environment were constructed.The technical solution of using USB flash drive to store the ported content was proposed,and the specific porting operation instruc-tions and initialization scripts were given.Experiments were carried out in Thinkpad firmware.Result show that the transplanted deep learning framework can run normally in image classification,natural language processing,malicious code detection and malicious code classification,and performs well in various evaluation indexes.
关键词
操作系统/开源固件/深度学习/深度学习框架移植/图像分类/自然语言处理/恶意代码检测/恶意代码分类Key words
operating system/open-source firmware/deep learning/deep learning framework migration/image classification/natural language processing/malicious code detection/malicious code classification引用本文复制引用
出版年
2024
NETL
NSTL
万方数据