Privacy of Machine Learning

Update: We decide to cancel the presentation on 24.01, please check the new schedule.

-------------------------------------------------------------

 

Dear all,

After receiving your responses, we have arranged a schedule for you to give the presentations (see it at the end of this message).

Update: Next Monday is a holiday, there will be no seminar.

Thus, the presentation will start on 08.11.  Every Monday from 2 pm to 4 pm, we will have two presenters introduce their preferred papers.

See you next week. :)

Best,
Rui

 

-----------------------------------------------------------------

08.11:

1. Xinyue Shen, ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Models

2. Yiyong Liu, Membership Leakage in Label-Only Exposures

 

15.11:

3. Xiangyu Dong, Extracting Training Data from Large Language Models

4. Elisa Ebler, Privacy Risks of Securing Machine Learning Models against Adversarial Examples

 

22.11:

5. Kazi Fozle Azim Rabi, Practical Blind Membership Inference Attack via Differential Comparison

6. Zeyang Sha, Quantifying and Mitigating Privacy Risks of Contrastive Learning

 

29.11:

7. Yixin Wu, Overlearning Reveals Sensitive Attributes

8. Ziqing Yang, Auditing Data Provenance in Text-Generation Models

 

06.12:

9. Tanvi Ajay Gunjal, The Secret Sharer: Evaluating and Testing Unintended Memorization in Neural Networks

10. Minxing Zhang, Membership Inference Attacks Against Recommender Systems

 

13.12:

11. Yiting Qu, Quantifying Privacy Leakage in Graph Embedding

12. Kavu Maithri Rao, The Secret Revealer: Generative Model-Inversion Attacks Against Deep Neural Networks.

 

10.01:

13. Prajvi Saxena, Updates-Leak: Data Set Inference and Reconstruction Attacks in Online Learning

14. MANUELA CERON, When Machine Unlearning Jeopardizes Privacy

 

17.01:

15. Elisa Ebler, Privacy Risk in Machine Learning: Analyzing the Connection to Overfitting

16. Ankita Behura, Membership Inference Attacks Against Machine Learning Models

 

 

Dear all,

The paper list is online, please select your favorite paper and mark its index on the doodle (https://doodle.com/poll/en9y69hepds8wg9h?utm_source=poll&utm_medium=link) by tomorrow 4 pm. One important change, no one needs to choose three papers anymore (that's what Yang mentioned half an hour ago), instead, we will use the following rules.
For those who participate as pro-seminar students (Wang, Ye; Ebler, Elisa; and O Keefe, Shannon Angela), please select two papers that you want to present. For other students, please select one paper. The assignment will be based on the first-come, first-serve principle. The paper selection deadline is Tuesday (26.10) 4 pm.
If you want to present the paper next week (01.11), you will get extra points, please contact me (rui.wen@cispa.de) if you are interested.

Best,
Rui