Artificial Intelligence Model Based Security Protection Method for IoT Applications

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.