Security and Privacy in Data-Intensive, High-Performance Computing Contexts


October 2, 2017
2:00pm to 3:00pm
190 Doe Library
Get Directions

Computer security in unclassified, scientific research environments can easily be dismissed as unimportant. However, scientific research can also be subject to privacy considerations and regulations (e.g., biomedical data, social science data, smart grid data, transportation data), be of significant intellectual property value (e.g., commercially viable), and/or have some bearing on national security (e.g., export controlled, official use only).

Traditionally, security solutions in scientific research environments have been to throw up air gaps or big firewalls to "silo" such science. But air gaps and firewalls are antithetical to modern, data-intensive collaborative science, which seeks to enable data computation and movement that is as fast as possible, and data sharing that is as flexible as possible, all while maintaining security and privacy protections that institutional security personnel, government regulators, and others, would find to be equivalent to or stronger than "traditional" approaches. High-performance methods for keeping data secret, such as fully-homomorphic techniques to compute or searching over encrypted data, are needed, as well as high-performance methods for selectively disclosing private information, such as secure, multi-party computation and differential privacy.

This talk and discussion seeks to explore current state of and near-future possibilities of techniques at the intersection of high-performance, data-intensive science, and cutting edge security and privacy techniques to protect scientific data. The goal of this talk is to develop dialogues and partnerships with domain scientists with science relevant to these goals, and computer scientists interested in developing techniques that might help achieve these goals, with the goal of eventually deploying such techniques in high-performance computing and networking environment, such as LBNL's NERSC supercomputing facility, and LBNL's ESnet research network backbone.

The BIDS Open Table series features brief (10-minute) introductory research talks and active discussion sessions in an informal atmosphere designed to facilitate new collaborations, especially among campus researchers and disciplines new to the BIDS community. All are welcome and encouraged to attend.


Sean Peisert

Staff Scientist, Computational Research Division
Lawrence Berkeley National Laboratory

Dr. Sean Peisert is jointly appointed as a staff scientist at Lawrence Berkeley National Laboratory; as chief cybersecurity strategist for CENIC; and as an associate adjunct professor of Computer Science at the University of California, Davis. One of his key interests is developing distributed, high-performance computer security techniques that enable scientific research that is otherwise constrained by technical limitations. At LBNL, he works closely with NERSC, LBNL's supercomputing facility, and ESnet, LBNL's 100G, high-throughput scientific network. At CENIC, he is responsible for cybersecurity strategy and implementation for CENIC's enterprise as well as for CalREN, a high-capacity network designed to meet the unique requirements of CENIC's constituent population of over 20 million users, including all of the University of California, Stanford, Caltech, and USC, including all of the academic medical centers.