AI-Based Anonymization Techniques for Healthcare Data

EasyChair Preprint 12694

Abstract

Healthcare data contains sensitive and personally identifiable information (PII), necessitating the use of effective anonymization techniques to protect patient privacy. With the advancements in artificial intelligence (AI), novel AI-based anonymization techniques have emerged, offering innovative approaches to address privacy and security concerns in healthcare data sharing. This paper provides an overview of AI-based anonymization techniques for healthcare data.

The traditional anonymization techniques, such as de-identification, aggregation, and data masking, are discussed, highlighting their limitations and challenges. The role of AI in enhancing these techniques is explored, showcasing the potential of AI for improving accuracy and efficiency in PII detection and synthetic data generation.

Furthermore, the paper delves into advanced AI-based anonymization techniques, including differential privacy and homomorphic encryption. These techniques enable privacy preservation while maintaining data utility by adding controlled noise to data or performing computations on encrypted data.

Critical aspects of evaluating and validating AI-anonymization techniques are discussed, emphasizing the assessment of re-identification risks and data utility, as well as compliance with regulations such as HIPAA and GDPR. Challenges and limitations associated with adversarial attacks, privacy-utility trade-offs, and ethical considerations are also addressed..

Keyphrases: AI-based anonymization, Data Encryption, Data Masking, Healthcare data, Personally Identifiable Information (PII), Privacy, Synthetic Data Generation, aggregation, de-identification, differential privacy, machine learning

@booklet{EasyChair:12694,
  author    = {Abdul Jawwad},
  title     = {AI-Based Anonymization Techniques for Healthcare Data},
  howpublished = {EasyChair Preprint 12694},
  year      = {EasyChair, 2024}}