Privacy-preserving methods: Building secure projects
summary
Conference notes state privacy laws grow from 20 to 100, and the session shows concrete techniques: randomized response gives plausible deniability for sensitive surveys, differential privacy adds calibrated noise with an explicit epsilon and a privacy budget to protect individuals while preserving aggregate statistics, federated learning trains models on-device but still leaks information via model deltas, and homomorphic encryption enables remote computation on ciphertexts but remains slow and hard to scale for large datasets.
AI-written summary · about
chapters (8)
- 0:00Intro
- 0:23Privacy-preserving methods: Building secure projects
- 3:34You want to collect and release data that contains answers to sensitive questions
- 7:06You want to collect and release data that eontains answers te sensitive questions
- 11:51You want to make generalizations over a population
- 13:35In the context of a database Given we perform some query on the database we remove a person from the database and the query does not change then that person's privacy is fully protected
- 17:22You want to use prediction models with user's data
- 20:37You want to update your model with user's data
related talks
37mDemystifying Privacy Preserving Computing
16mPrivacy by Design for JavaScript Data Systems
42mPrivacy Challenges in Machine Learning
39mPrivacy-preserving Data Sharing
38mDemystifying Privacy Preserving Computing
01h 17mTutorial: Privacy, Part I
58mFormal Methods and Proofs of Privacy Properties, Part I
33mPractical Data Privacy