John Mitchell: 2014 Security Workshop


Monday, April 14, 2014
Location: Fisher Conference Center, Arrillaga Alumni Center

"Machine-Learning-Based Data Compression and Secure Oblivious Computation"


In one approach to secure cloud computing, the cloud processor runs a secrecy-preserving computation on encrypted data. Such computation must be data-oblivious, meaning that the control flow of the program cannot depend on input data. One reason that data-oblivious algorithms can be prohibitively inefficient is that encrypted data must be accessed according to the same pattern for any input. Working with the visually compelling example of data-oblivious routing on street maps, we show how machine learning techniques can be used to compress a street map into a compact representation that can be efficiently decoded by a data-oblivious algorithm. Preliminary results indicate that the system may be efficient enough to enable practical, privacy-preserving directions and navigation services.

This is joint work with Joe Zimmerman and Jeremy Planul.


John Mitchell is Professor of Computer Science, Vice Provost for Online Learning, and the Mary and Gordon Crary Family Professor in the School of Engineering. His organization on campus supports the instructional design, production and delivery of online teaching and learning material. As a result of his seed grant programs, Stanford funded over 30 experimental faculty activities in 2012 to develop and deploy new online activities in on-campus or externally available courses. As a professor of computer science, Mitchell's research interests include computer security, privacy, programming languages, mathematical logic, and web technology.