38 lines
1.5 KiB
Markdown
38 lines
1.5 KiB
Markdown
|
Security and Privacy are emerging as very important research areas.
|
||
|
|
Vulnerabilities in software are found and exploited almost everyday
|
||
|
|
and with disastrous consequences (e.g., the Equifax massive data
|
||
|
|
breach). Moreover, our private data is increasingly at risk and thus
|
||
|
|
techniques that enhance privacy of sensitive data (known as
|
||
|
|
privacy-enhancing technologies (PETS)) are becoming increasingly
|
||
|
|
important. Also, machine-learning (ML) is increasingly being utilized to
|
||
|
|
make decisions in critical sectors (e.g., health care, automation, and
|
||
|
|
finance). However, in deploying these algorithms presence of malicious
|
||
|
|
adversaries is generally ignored.
|
||
|
|
|
||
|
|
This advanced topics class will tackle techniques related to all these
|
||
|
|
themes. We will investigate techniques to make software more secure.
|
||
|
|
Techniques for ensuring privacy of sensitive data will also be
|
||
|
|
covered. Adversarial ML (what happens to ML algorithms in the
|
||
|
|
presence of adversaries?) will be also be discussed. List of some
|
||
|
|
topics that we will cover (obviously not complete) are given below.
|
||
|
|
|
||
|
|
Software Security:
|
||
|
|
- Information flow
|
||
|
|
- Techniques for finding vulnerabilities in software
|
||
|
|
- Defense techniques (e.g., control-flow integrity)
|
||
|
|
|
||
|
|
Privacy:
|
||
|
|
- Differential Privacy
|
||
|
|
- Zero-knowledge proofs
|
||
|
|
- Secure multi-party computation
|
||
|
|
|
||
|
|
Adversarial ML:
|
||
|
|
- Training-time attacks
|
||
|
|
- Test-time attacks
|
||
|
|
- Model Theft attacks
|
||
|
|
|
||
|
|
Grading: There are three components that relate to grading:
|
||
|
|
- Reading research papers and writing reviews.
|
||
|
|
- Few homeworks.
|
||
|
|
- Class project.
|