Privacy and Security


Data Privacy

The basic requirement is that the owners of the data should

  • know
    • what is collected
    • what it will be used for
  • be able to
    • view it
    • ask for any part of it to be corrected or deleted completely.

Companies that use use the data

  • are responsible for its safekeeping
  • should not share it without the authorization of the original owner

Data Sanitizer

Personal data should be sanitized by hiding all the personal details.

  • Anonymization is not enough
    • e.g. Fileds such as birth date and zip code are informative

Data Diversity

  • Machine learning programs are trained on data collected in the past, when there may have been lack of diversity
    • e.g. a bank may have had more men than women as customers in the past

Explainable Artifical Intelligence (XAI)

  • We want models to be written down in a form understandable by human.
  • The experts in that area can access and analyze the model.
Last modified March 10, 2021