Ensuring Privacy and Security in Machine Learning: A Novel Approach to Efficient Data Removal

View Sample PDF

Author(s): Velammal (Anna University, Chennai, India)and N. Aarthy (Anna University, Chennai, India)
Copyright: 2024
Pages: 18
Source title: The Ethical Frontier of AI and Data Analysis
Source Author(s)/Editor(s): Rajeev Kumar (Moradabad Institute of Technology, India), Ankush Joshi (COER University, Roorkee, India), Hari Om Sharan (Rama University, Kanpur, India), Sheng-Lung Peng (College of Innovative Design and Management, National Taipei University of Business, Taiwan)and Chetan R. Dudhagara (Anand Agricultural University, India)
DOI: 10.4018/979-8-3693-2964-1.ch009

Keywords: Artificial Intelligence / Ethics & Law / Information Science Reference / Security & Forensics

Purchase

View Ensuring Privacy and Security in Machine Learning: A Novel Approach to Efficient Data Removal on the publisher's website for pricing and purchasing information.

Abstract

Modern systems generate vast amounts of data, creating complex data networks. Users prioritize the safety, security, and privacy of their data. This project focuses on efficiently removing or erasing data from the machine learning model upon user request, addressing privacy concerns. Under GDPR, users can request the deletion of sensitive data from both user records and the machine learning model that has processed the data. Additionally, the project employs the SISA approach to address errors and attacks by dividing the dataset into shards and implementing a slice-based ensemble learning technique. Each shard functions as an independent model, and after training, a majority voting approach aggregates these models into a final model. Experimental results demonstrate reduced retraining costs, as only the remaining slices are retrained instead of the entire model.

The IRMA Community

Research IRM

Ensuring Privacy and Security in Machine Learning: A Novel Approach to Efficient Data Removal

Purchase

Abstract

Related Content

IRMA Sponsors