Research Article
Artificial Intelligence Model Based Security Protection Method for IoT Applications
@INPROCEEDINGS{10.1007/978-3-031-65126-7_15, author={Xiaolong Luo and Xiaoli Chen and Jie Wei and Liang Zhang and Luping Xu and Bijun Zhao}, title={Artificial Intelligence Model Based Security Protection Method for IoT Applications}, proceedings={Quality, Reliability, Security and Robustness in Heterogeneous Systems. 19th EAI International Conference, QShine 2023, Shenzhen, China, October 8 -- 9, 2023, Proceedings, Part I}, proceedings_a={QSHINE}, year={2024}, month={8}, keywords={Model privacy security internet of things artificial intelligence model encryption technology access control}, doi={10.1007/978-3-031-65126-7_15} }- Xiaolong Luo
Xiaoli Chen
Jie Wei
Liang Zhang
Luping Xu
Bijun Zhao
Year: 2024
Artificial Intelligence Model Based Security Protection Method for IoT Applications
QSHINE
Springer
DOI: 10.1007/978-3-031-65126-7_15
Abstract
The issue of model privacy security is increasingly affecting the application systems of Artificial Intelligence Internet of Things (AI-IOT) terminals, where it is challenging to protect the privacy of the underlying AI models. In this paper, we propose a security protection RC6-plus algorithm based on cryptography and access control for AI model security in IoT applications. Specifically, the proposed method effectively protects the privacy of crucial algorithms in the program by encrypted storing the model parameters, as well as storing and code obfuscating the neural network structure and parameters of the AI model independently while adding the isolation treatment of the JNI communication layer. The results of the experiments verify the effectiveness of the proposed method.
