layout | title | nav_exclude | permalink | seo | ||||
---|---|---|---|---|---|---|---|---|
home |
CSE 467 |
true |
/:path/ |
|
With the rise of data-driven technologies, including machine learning and artificial intelligence-based systems, comes enormous risks to protect the data from unauthorized and inappropriate use. The data comes from our everyday use of digital technologies, as they store all of our interactions, and learn from them. However, that data also reveals intimate details about ourselves, our lives, and our daily activities. Such knowledge is highly privacy sensitive and can be used by unauthorized entities if appropriate safeguarding mechanisms are not in place. Additionally, ML-based models, which are trained on those data, are vulnerable to adversarial attacks. This course will equip students with theoretical and hands-on knowledge on how to protect private data, and other technologies that are built on those data.
After the end of the semester, you will learn
- Different definitions of data privacy and security
- Computational mechanisms to protect data from unauthorized access and use
- Privacy and security attacks on machine learning-based models and how to prevent them
- How to build robust machine learning models using a minimal set of features
- How to incorporate security and privacy into data analytics and machine learning models while retaining data utility
- Hands-on experience with implementing privacy and security algorithms using popular open-source libraries
- In-class quizzes
- 3 to 4 Homeworks
- One Term project
- Basics Probability Theory
- Algorithms and Data structures
- Basics of database systems
- Python/R programming language (or willingness to learn on their own)
- Privacy threat modeling
- Identification and inference attacks
- De-identification mechanisms
- Data utility and privacy trade-offs.
- Basics of Differential Privacy
- Basics of Predictive Machine Learning
- Private Information Disclosure by ML models
- Basics of Federated Machine Learning
- Differential Privacy with Federated Learning
- Secure and Private implementations of ML models
- Causal mechanisms for privacy-preserving machine learning
- Usability issues in security and privacy engineering
- Privacy and security issues that are unsolvable by technology alone.