Privacy-Preserving Decision Tree Classification Using Homomorphic Encryption in IoT Big Data Scenarios

To effectively address the issue of privacy-preserving decision tree classification services in IoT big data scenarios, this study combines decision tree classification models with homomorphic encryption techniques and proposes an efficient protocol for privacy-preserving decision tree classification. The protocol consists of three stages: encryption and obfuscation of the decision tree classification model, privacy-preserving comparison using homomorphic computations, and secure retrieval of classification results. The protocol ensures the confidentiality of both the service provider's decision tree model parameters and structure, as well as the user's feature data used for classification. Security analysis demonstrates that the proposed protocol is resistant to attacks from “honest-but-curious” adversaries. Experimental evaluations, conducted using decision tree models trained on public datasets, compare the proposed protocol against existing methods in terms of accuracy, classification, and efficiency time the service classification, preserving privacy. Results confirm the accuracy and efficiency of the proposed protocol.