• Skip to main content
  • Skip to primary navigation
  • Departments
    • Bioengineering
    • Civil and Environmental Engineering
    • Electrical Engineering and Computer Sciences
    • Industrial Engineering and Operations Research
    • Materials Science and Engineering
    • Mechanical Engineering
    • Nuclear Engineering
    • Aerospace program
    • Engineering Science program
  • News
    • Berkeley Engineer magazine
    • Social media
    • News videos
    • News digest (email)
    • Press kit
  • Events
    • Events calendar
    • Commencement
    • Homecoming
    • Cal Day
    • View from the Top
    • Kuh Lecture Series
    • Minner Lecture
  • College directory
  • For staff & faculty
Berkeley Engineering

Educating leaders. Creating knowledge. Serving society.

  • About
    • Facts & figures
    • Rankings
    • Mission & values
    • Equity & inclusion
    • Voices of Berkeley Engineering
    • Leadership team
    • Milestones
    • Facilities
    • Maps
  • Admissions
    • Undergraduate admissions
    • Graduate admissions
    • New students
    • Visit
    • Maps
    • Admissions events
    • K-12 outreach
  • Academics
    • Undergraduate programs
    • Majors & minors
    • Undergraduate Guide
    • Graduate programs
    • Graduate Guide
    • Innovation & entrepreneurship
    • Kresge Engineering Library
    • International programs
    • Executive education
  • Students
    • New students
    • Advising & counseling
    • ESS programs
    • CAEE academic support
    • Student life
    • Wellness & inclusion
    • Undergraduate Guide
    • > Degree requirements
    • > Policies & procedures
    • Forms & petitions
    • Resources
  • Research & faculty
    • Centers & institutes
    • Undergrad research
    • Faculty
  • Connect
    • Alumni
    • Industry
    • Give
    • Stay in touch
Home > News > In training
Raluca Ada Popa and Jean-Luc WatsonRaluca Ada Popa and Jean-Luc Watson (Photos by Adam Lau/Berkeley Engineering and courtesy Jean-Luc Watson)

In training

Berkeley Engineer magazine Spring 2023
May 17, 2023
This article appeared in Berkeley Engineer magazine, Spring 2023
  • In this issue
    Illustration of blood cells

    Forever young

    Scott Moura holding an electric car charger

    Smart and in charge

    Prepping the Eureka-1 rocket for launch

    A Eureka-1 moment

    Students watching model airplane in wind tunnel

    Engineering new frontiers

    Upfront

    • Raising the stakes
    • Life isn't fair ... yet
    • Working like a dog
    • In training
    • As tough as they come
    • Speak and spell
    • Keep it cool
    • Stopping traffic

    New & noteworthy

    • Distinguished Teaching Awards
    • On the grid
    • With flexibility comes possibility
    • Farewell
  • Past issues

Keeping sensitive data safe has sometimes come at the expense of speed when training machines to perform automated tasks like biometric authentication and financial fraud detection. But no longer. Raluca Ada Popa, associate professor of electrical engineering and computer sciences, and Ph.D. student Jean-Luc Watson have developed an innovative privacy-preserving approach to machine learning using a new platform, dubbed Piranha, that harnesses the speed of graphics processing units (GPUs) to train a realistic neural network on encrypted data for the first time.

“Even though people have wanted to do this for at least 20 years, training a realistic neural network model while keeping the data encrypted has not been practical,” said Popa. “The key was to make GPUs work with encrypted computation.”

GPUs can process large amounts of data simultaneously, making them ideal for high-performance computing and deep-learning applications. While they can be used to quickly train neural networks on plain text, they do not work with encrypted data. Encrypted data is incompatible with GPUs because it uses integers instead of floats — another kind of numerical data — and accesses memory in non-standard ways.

Piranha addresses these issues with a three-layer architecture that allows applications to interoperate with any cryptographic protocol. The researchers, including postdoctoral researcher Sameer Wagh, showed that they could train a realistic neural network, end to end, on encrypted data in a little over a day, a significant performance gain over previous approaches. They estimated that accomplishing the same task on Falcon, a state-of-the-art predecessor to Piranha, would have required 14 days, making it prohibitively expensive and impractical.

“With Piranha, we not only trained a realistic network for the first time with encrypted data, but we also improved performance by 16 to 48 times,” said Popa.

Learn more: For your eyes only; Piranha: A GPU platform for secure computation (USENIX)

Topics: Security & privacy, Computer science, Computing
  • Contact
  • Give
  • Privacy
  • UC Berkeley
  • Accessibility
  • Nondiscrimination
  • instagram
  • X logo
  • linkedin
  • facebook
  • youtube
© 2025 UC Regents