1 / 22

Multiparty Computation and Application

EPIC is a groundbreaking solution that enables efficient and private image classification through secure multiparty computation. It ensures active security, surpassing other protocols in related work, and achieves impressive performance gains over state-of-the-art systems. EPIC provides faster runtime, significant improvement in communication cost, and higher classification accuracy. Explore how EPIC can transform image classification with enhanced security, efficiency, and accuracy.

Download Presentation

Multiparty Computation and Application

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Multiparty Computation and Application CRYP-F01 EleftheriaMakri Lecturer/Researcher Imec-COSIC, KU Leuven, BE & Saxion University of Applied Sciences, NL @MakriEleftheria

  2. EPIC: Efficient Private Image Classification (or: Learning from the Masters) E. Makri, D. Rotaru, N. P. Smart, F. Vercauteren

  3. EPIC: Efficient Private Image Classification

  4. Feature Extraction

  5. Transfer Learning Feature Extraction (or: Learning from the Masters)

  6. EPIC Security Active Security vs. Passive Security All other protocols in the related work EPIC

  7. Step 1: Create the ML model Alice Inception–v3 CNN Linear SVM Dragos Rotaru 9

  8. Step 2: Alice secret shares the ML model Charlie Alice Bob

  9. Step 2: Alice secret shares the ML model Charlie Alice Bob Dragos Rotaru + + =

  10. Step 3: Bob extracts features Bob Inception–v3 CNN Features Dragos Rotaru 12

  11. Step 4: Bob secret shares features Charlie Alice Bob Dragos Rotaru

  12. Step 4: Bob secret shares features Charlie Alice Bob Dragos Rotaru Dragos Rotaru + + = CNN-Feat( )

  13. Step 5: Parties use MPC to help Charlie compute label of SVM-Alice(Bob-Image) Charlie Alice Bob

  14. Step 5: Parties use MPC to help Charlie compute label of SVM-Alice(Bob-Image) Charlie Alice ***“Florist”*** Bob

  15. Video of the Demo of our work will appear here

  16. EPIC Performance – Simple Variant Computation Cost Communication Cost

  17. Performance of the state-of-the-art private image classification Computation Cost Communication Cost * JianLiu, MikaJuuti, YaoLu, N. Asokan. ObliviousNeural Network Predictionsvia MiniONNTransformations. In Proceedings of the 2017 ACM SIGSAC Conference on Computer and Communications Security (pp. 619-631). ACM. ** ChiraagJuvekar, VinodVaikuntanathan, andAnanthaChandrakasan. GAZELLE: A low latencyframeworkfor secure neuralnetworkinference. In 27th USENIX Security Symposium (USENIX Security '18), Baltimore, MD, 2018. USENIX Association.

  18. EPIC Efficiency Gain over the state-of-the-art 1 ChiraagJuvekar, VinodVaikuntanathan, andAnanthaChandrakasan. GAZELLE: A low latencyframeworkfor secure neuralnetworkinference. In 27th USENIX Security Symposium (USENIX Security '18), Baltimore, MD, 2018. USENIX Association. • EPIC vs. Gazelle1 on CIFAR-10: • 34 times faster runtime; • 50 times improvement of communication cost; • 7% higher classification accuracy. • EPIC vs. Gazelle1 with the same accuracy: • 700 times faster runtime; • 500 times improvement of communication cost.

  19. Now what? • What would transform EPIC to a LEGENDARY solution? • Maintain security • Maintain or increase efficiency • Increase accuracy! • Any ideas on how to do this (using MPC)? • Talk to me during the break, or • Contact me offline at: eleftheria.makri@esat.kuleuven.be

  20. THAT WAS EPIC!

More Related