DP-SMOTE: Integrating Differential Privacy and Oversampling Technique to Preserve Privacy in Smart Homes

Elsayed, Amr; Alsharkawy, Almohammady; Farag, Mohamed; Abu Yusuf, Shaban

doi:10.58675/2636-3305.1669

	DP-SMOTE: Integrating Differential Privacy and Oversampling Technique to Preserve Privacy in Smart Homes
Al-Azhar Bulletin of Science
Volume 2024, Issue 1, January 2024
DOI: 10.58675/2636-3305.1669
View on SCiNiTO
Authors
Amr Elsayed¹; Almohammady Alsharkawy¹; Mohamed Farag²; Shaban Abu Yusuf³
¹Computer Science Department, Faculty of Science (Boys), Al-Azhar University, Cairo, Egypt
²Computer Science Department, Obour High Institute for Informatics, Cairo, Egypt
³Department of Mathematics, Faculty of Science (Boys), Al-Azhar University, Cairo, Egypt
Abstract
Smart homes represent intelligent environments where interconnected devices gather information, enhancing users’ living experiences by ensuring comfort, safety, and efficient energy management. To enhance the quality of life, companies in the smart device industry collect user data, including activities, preferences, and power consumption. However, sharing such data necessitates privacy-preserving practices. This paper introduces a robust method for secure sharing of data to service providers, grounded in differential privacy (DP). This empowers smart home residents to contribute usage statistics while safeguarding their privacy. The approach incorporates the Synthetic Minority Oversampling technique (SMOTe) and seamlessly integrates Gaussian noise to generate synthetic data, enabling data and statistics sharing while preserving individual privacy. The proposed method employs the SMOTe algorithm and applies Gaussian noise to generate data. Subsequently, it employs a k-anonymity function to assess re-identification risk before sharing the data. The simulation outcomes demonstrate that our method delivers strong performance in safeguarding privacy and in accuracy, recall, and f-measure metrics. This approach is particularly effective in smart homes, offering substantial utility in privacy at a re-identification risk of 30%, with Gaussian noise set to 0.3, SMOTe at 500%, and the application of a k-anonymity function with k ¼ 2. Additionally, it shows a high classification accuracy, ranging from 90% to 98%, across various classification techniques.


Statistics Article View: 26