Scholarlink

<p>The need for real-time processing of educational data versus the sensitivity of student data makes the balancing between privacy and accessibility a vital challenge in data analysis. We present the PSEDA model (Privacy-Secure Educational Data Analysis) that integrates homomorphic encryption, differential privacy and role-based access control for strong privacy protection while permitting analysis. The primary breakthrough of the model is its ability to process large volumes of information without incurring huge costs due to encryption and decryption. We consistently outperform the state-of-the-art over the tasks within PSEDA. It obtains the highest accuracy of 96.7% at 30 features, but also provides the lowest time cost for encryption 1009&#xa0;ms and decryption 704&#xa0;ms at 5 features. Furthermore, it achieves an impressive 90.8% for access restriction and shows a very high throughput of 90.4% at 5 features. Based on assessment on the EdNet dataset, these findings demonstrate that the PSEDA model can securely and efficiently operate on large-scale educational data, reflecting a significant step forward in privacy-preserving data analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Privacy-enhanced secure framework for educational data protection and analysis

  • 期刊论文

  • 发表日期：

    2025年3月17日

查看全文
  • V. Balachandar,
  • K. Venkatesh

摘要

The need for real-time processing of educational data versus the sensitivity of student data makes the balancing between privacy and accessibility a vital challenge in data analysis. We present the PSEDA model (Privacy-Secure Educational Data Analysis) that integrates homomorphic encryption, differential privacy and role-based access control for strong privacy protection while permitting analysis. The primary breakthrough of the model is its ability to process large volumes of information without incurring huge costs due to encryption and decryption. We consistently outperform the state-of-the-art over the tasks within PSEDA. It obtains the highest accuracy of 96.7% at 30 features, but also provides the lowest time cost for encryption 1009 ms and decryption 704 ms at 5 features. Furthermore, it achieves an impressive 90.8% for access restriction and shows a very high throughput of 90.4% at 5 features. Based on assessment on the EdNet dataset, these findings demonstrate that the PSEDA model can securely and efficiently operate on large-scale educational data, reflecting a significant step forward in privacy-preserving data analysis.