The 2nd International Workshop on Federated Learning for Data Privacy and Confidentiality (in Conjunction with NeurIPS 2019)

FL - NeurIPS 2019


Artificial Intelligence



Overview
Privacy and security are becoming major concerns in recent years, particularly as companies and organizations are collecting increasingly detailed information about their products and users. This information can enable machine learning that produces more helpful products. However, at the same time, it expands the potential for misuse, and increases corresponding public concerns about the way companies use data, particularly when private data about individuals is involved. Recent research shows that privacy and utility do not necessarily need to be at odds, but can be addressed by careful design and analysis. The need for such research is reinforced by the recent introduction of new legal constraints, led by the European Union’s General Data Protection Regulation (GDPR), which is already inspiring novel legislative approaches around the world such as Cyber-security Law of the People’s Republic of China and The California Consumer Privacy Act of 2018.
A specific approach that has the potential to address a number of problems in this space is Federated Learning. The concept of Federated Learning is relevant in the setting when one wants to train a machine learning model based on a dataset stored across multiple locations, without the ability to move the data to any central location. This seemingly mild restriction renders many of the state-of-the-art techniques in machine learning impractical. One class of applications arises when data is generated by different users of a smartphone app, staying on users’ phones for privacy reasons. For example, Google’s Gboard mobile keyboard is already using federated learning in multiple places. Another class of applications involves data collected by different organizations, unable to share due to confidentiality reasons. Nevertheless, the same restrictions can also be present independent of privacy concerns, such as the case of data streams collected by IoT devices or self-driving cars, which need to be processed on-device, because it is infeasible to transmit and store the sheer amount of data.
At this moment, the pace of research innovation in federated learning is hampered by the relative complexity of properly setting up even simple experiments that reflect the practical setting. This issue is exacerbated in academic settings which typically lack access to actual user data. Recently, multiple open-source projects were created to address this high-barrier to entry. For example, LeaF is a benchmarking framework that contains preprocessed datasets, each with a “natural” partitioning that aims to reflect the type of non-identically distributed data partitions encountered in practical federated environments. Federated AI Technology Enabler (FATE) led by WeBank is an open-source technical framework that enables distributed and scalable secure computation protocols based on homomorphic encryption and multi-party computation, supporting federated learning architectures with various machine learning algorithms. Webank is also leading a related IEEE standard proposal. TensorFlow Federated (TFF) led by Google is an open-source framework on top of TensorFlow for flexibly expressing arbitrary computation on decentralized data. TFF enables researchers to experiment with federated learning on their own datasets, or those provided by LeaF. Google has also published a systems paper describing the design of their production system, which supports tens of millions of mobile phones. We expect these projects will encourage academic researchers and industry engineers to work more closely in addressing the challenges and eventually make significant positive impact. We support reproducible research and will sponsor a prize to be given to the best contribution, which also provides code to reproduce their results.
The workshop aims to bring together academic researchers and industry practitioners with common interests in this domain. For industry participants, we intend to create a forum to communicate what kind of problems are practically relevant. For academic participants, we hope to make it easier to become productive in this area. Overall, the workshop should provide an opportunity to share the most recent and innovative work in this area, and discuss open problems and relevant approaches. The technical issues encouraged to be submitted include general computation based on decentralized data (i.e., not only machine learning), and how such computations can be combined with other research fields, such as differential privacy, secure multi-party computation, computational efficiency, coding theory, and others. Contributions in theory as well as applications are welcome, particularly proposals for novel system design.
Call for Contributions
We welcome high quality submissions in the broad area of federated learning (FL). A few (non-exhaustive) topics of interest include:
Optimization algorithms for FL, particularly communication-efficient algorithms tolerant of non-IID data
Approaches that scale FL to larger models, including model and gradient compression techniques
Novel applications of FL
Theory for FL
Approaches to enhancing the security and privacy of FL, including cryptographic techniques and differential privacy
Bias and fairness in the FL setting
Attacks on FL including model poisoning, and corresponding defenses
Incentive mechanisms for FL
Software and systems for FL
Novel applications of techniques from other fields to the FL setting: information theory, multi-task learning, model-agnostic meta-learning, and etc.
Work on fully-decentralized (peer-to-peer) learning will also be considered, as there is significant overlap in both interest and techniques with FL.
Submissions in the form of extended abstracts must be at most 4 pages long (not including references) and adhere to the NeurIPS 2019 format. Submissions should be anonymized. The workshop will not have formal proceedings, but the accepted contributions will be expected to present a poster at the workshop.
Submission link:
https://easychair.org/conferences/?conf=flneurips2019
Organizers
Lixin Fan
Jakub Konečný
Yang Liu
Brendan McMahan
Virginia Smith
Han Yu