CVPR Tutorial On Distributed Private Machine Learning for Computer Vision: Federated Learning, Split Learning and Beyond

This tutorial presents different methods for protecting confidential data on clients while still allowing servers to train models. In particular, we focus on distributed deep learning approaches under the constraint that local data sources of clients (e.g. photos on phones or medical images at hospitals) are not allowed to be shared with the server or amongst other clients due to privacy, regulations or trust. We describe such methods that include federated learning, split learning, homomorphic encryption, and differential privacy for securely learning and inferring with neural networks. We also study their trade-offs with regards to computational resources and communication efficiency in addition to sharing practical know-how of deploying such systems.

Takeaways: Attendees will get an overview of secure, private distributed deep learning for images and videos. Hands on examples of different distributed deep learning techniques that operate while protecting confidential patterns in data will be shared. Finally, several in depth examples demonstrating how these techniques can be implemented across various configurations in different application domains will be shared.

Tutorial Schedule (in PM):
01:30 – 01:50 Module I: Introduction to Federated learning and Split learning
by Ramesh Raskar, MIT

01:50 – 02:45 Module II: Federated learning: Machine learning on decentralized data
by Brendan McMahan, Google

02:45 – 02:55 Q&A

02:55 – 03:40 Module III A: Split learning: Resource efficient distributed machine learning
by Otkrist Gupta, LendBuzz

Module III B: Split learning: Reducing leakage in distributed machine learning
by Praneeth Vepakomma, MIT

03:40 – 04:00 Q&A and Break

04:00 – 04:25 Module IV: Homomorphic Encryption for Neural Networks & Privacy Policies in Cloud
by Hassan Takabi, UNT

04:25 – 05:10 Module V: Federated learning at Google - Research
by Jakub Konečný, Google

05:10 – 05:30 Discussion of Open Problems

Duration: Half-day (4 hours)

Brendan McMahan (Google, USA)
Hassan Takabi (University of North Texas, Texas, USA)
Praneeth Vepakomma (MIT Media Lab, Cambridge, Massachusetts, USA)
Ramesh Raskar (MIT Media Lab, Cambridge, Massachusetts, USA)

Brendan McMahan (Google, USA)
Jakub Konečný (Google, USA)
Otkrist Gupta (LendBuzz)
Ramesh Raskar (MIT Media Lab, Cambridge, Massachusetts, USA)
Hassan Takabi (University of North Texas, Texas, USA)
Praneeth Vepakomma (MIT Media Lab, Cambridge, Massachusetts, USA)

The key papers from the authors are:
1.) (Federated learning) Federated learning: Strategies for improving communication efficiency,
Jakub Konečný, Brendan McMahan, Felix X. Yu, Ananda Theertha Suresh & Dave Bacon, PDF (2016).

2.) (Federated learning) Towards federated learning at scale: System design,
Jakub Konečný, Keith Bonawitz, Hubert Eichner, Wolfgang Grieskamp, 2nd SysML conference, PDF, (2019).

3.) (Split learning) Distributed learning of deep neural network over multiple agents,
Otkrist Gupta and Ramesh Raskar, Journal of Network and Computer Applications, PDF (2018)

4.) (Split learning) Reducing leakage in distributed deep learning for sensitive health data,
Praneeth Vepakomma, Otkrist Gupta, Abhimanyu Dubey, Ramesh Raskar, ICLR 2019 Workshop on AI for social good, PDF (2019).

5.) (Split learning) Split learning for health: Distributed deep learning without sharing raw patient data,
Praneeth Vepakomma, Otkrist Gupta, Tristan Swedish, Ramesh Raskar, ICLR 2019 Workshop on AI for social good, PDF (2018).

6.) (Homomorphic encryption for deep learning) CryptoDL: Deep Neural Networks over Encrypted Data,
Ehsan Hesamifard, Hassan Takabi, and Mehdi Ghasemi, PDF (2017).

7.) (Survey paper) No Peek: A Survey of private distributed deep learning,
Praneeth Vepakomma, Tristan Swedish, Ramesh Raskar, Otkrist Gupta, Abhimanyu Dubey, PDF (2018).

Visit project pages for ‘Federated Learning’ and ‘Split Learning’ for further details!