1 Privacy considerations in machine learning

 

This chapter covers

  • Concept and the importance of privacy protection in the Big Data AI era
  • Formulating different types of privacy-related threats, vulnerabilities, and attacks in machine learning, including reconstruction attacks, model inversion attacks, membership inference attacks, de-anonymization attacks, etc.
  • Presenting a summary of techniques that can be utilized in machine learning tasks in terms of minimizing and evading privacy risks and attacks

1.1 The Privacy Complications in the AI Era

1.2 The Threat of Learning Beyond the Intended Purpose

1.2.1 The Problem of Private Data in the Clear

1.2.2 Reconstruction Attacks

1.2.3 Model Inversion Attacks

1.2.4 Membership Inference Attacks

1.2.5 De-Anonymization or Re-Identification Attacks

1.2.6 Challenges of Privacy Protection in Big Data Analytics

1.3 Securing Privacy while Learning from Data: Privacy-Preserving Machine Learning

1.3.1 Use of Differential Privacy

1.3.2 Local Differential Privacy

1.3.3 Privacy-preserving Synthetic Data Generation

1.3.4 Privacy-preserving Data Mining Techniques

1.3.5 Compressive Privacy

1.4 How is This Book Structured?

1.5 Summary

sitemap