Privacy-preserving machine learning is an increasingly vital field as organizations and individuals strive to leverage machine learning while safeguarding sensitive data. This workshop addresses the growing demand for privacy-enhancing techniques. It offers an up-to-date overview of the key concepts, methods, and tools. Participants will be introduced to essential cryptographic techniques.

The workshop's structure consists of a keynote on Fully Homomorphic Encryption (FHE) by Andrei Stoian, a Data Scientist at Zama's Concrete-ML group. Additionally, it features contributions from various experts, including PhD students, early-stage researchers, and industry professionals.

The keynote delves into the practical aspects of FHE, discussing both advancements and challenges. Furthermore, the workshop invites contributions related to these topics, providing a platform for discussing real-world applications, enabling young researchers to showcase their work, and receive valuable insights and suggestions.

The PPML Conference welcomes contributed talks from PhD students, early-stage researchers, industrial and institutional experts, and experienced researchers in the field of Privacy-Preserving Machine Learning.

Submission Guidelines

If you have innovative research or insights to share, we invite you to submit your proposal for a contributed talk in one of the following topics for PPML or more.

All contributions must present results on Privacy Preserving techniques useful in the Machine Learning setting. A paper describing the contribution is not necessary but it can help to advance towards decision.

• A two pages abstract describing the contribution will be published in the proceedings of CIFRIS23.

Note that each talk presented at De Cifris workshops will be published according to a choice made by the paper authors as follows:

1. (either) A two-page mini-abstract will appear in the main Proceedings book,which will be the first issue of a new book series De Cifris Koine.The paper will also appear in other forms freely: since the mini-abstract is only two-page long, its publication in Koine will not affect the publication of the full paper in other venues.
2. (or) A two-page mini-abstract will appear in the main Proceedings book, which will be the first issue of a new book series De Cifris Koine and the paper will also appear in the first issue of (another) new book series Collectanea Ciphrarum. Papers appearing in Collectanea need to have been refereed, either by the workshop PC or by Collectanea’s board, and will have a typical length between 10 and 20 pages. Usually, a paper appearing in Collectanea will not be published in other venues (unless, possibly, with significant additional material).

List of Topics

• Differential Privacy
• Federated Learning
• Fully Homomorphic Encryption
• Secure Multiparty Computation
• Trusted Execution Environment

Program Committee & Organizing committee

• Michela Iezzi
• Marco Pedicini

Invited Speakers

• Andrei Stoian (Zama Concrete-ML, Paris)

Venue

The conference will be held in Rome at the same venue of CIFRIS23: at CONSOB HeadOffice (Via Giovanni Battista Martini, 3 - 00198 Rome, Italy)

Contact

All questions about submissions should be emailed to marco.pedicini@uniroma3.it

Questions about registration at the conference should be addressed to the cifris@decifris.it