Collaborative Learning with Obfuscation Assurances for grid Knowledge (CLOAK)

Project Summary

CLOAK (Collaborative Learning with Obfuscation Assurances for grid Knowledge) is a research project to investigate whether federated learning, combined with differential privacy (DP), allows electric utilities to train shared machine-learning models on sensitive operational data without releasing that data or accepting unquantified disclosure risk.

This project is supported by the U.S. Department of Energy’s Office of Cybersecurity, Energy Security, and Emergency Response (CESER) Risk Management Tools and Technologies (RMT) program.

Senior Personnel:

Sean Peisert (PI; LBNL)
Daniel Arnold (LLNL)
Jason Hollern (EPRI)
Kibaek Kim (ANL)
Jason MacDonald (LBNL)
Ashton Raffety (ANL)
Anna Scaglione (Cornell Tech)
Tong Wu (University of Central Florida)

Partners:

Argonne National Laboratory (ANL) Cornell Tech
Electric Power Research Institute (EPRI)
Lawrence Livermore National Laboratory (LLNL)
University of Central Florida (UCF)

More information is available on other Berkeley Lab R&D projects focusing on cybersecurity in general, as well as specifically on cybersecurity for energy delivery systems.