slide1
Download
Skip this Video
Download Presentation
Practical Private Computation of Vector Addition-Based Functions

Loading in 2 Seconds...

play fullscreen
1 / 8

Practical Private Computation of Vector Addition-Based Functions - PowerPoint PPT Presentation


  • 49 Views
  • Uploaded on

Yitao Duan and John Canny Computer Science Division University of California, Berkeley PODC 2007, August 12, Portland OR. Practical Private Computation of Vector Addition-Based Functions. Overview.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' Practical Private Computation of Vector Addition-Based Functions' - bianca


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.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
slide1
Yitao Duan and John Canny

Computer Science Division

University of California, Berkeley

PODC 2007, August 12, Portland OR

Practical Private Computation of Vector Addition-Based Functions
overview
Overview

A method for performing privacy preserving distributed computation of many algorithms that is practical and secure in a realistic threat model at large scale

Provably strong (information-theoretic) privacy

Efficient ZKP to deal with cheating users

model
Model
  • A few collaborating data miners mining data from n users
  • Each user has an m-dimensional vector
  • Realistic scale: m, n large (103–106)
  • Threat: data miners are passive, users are allowed to cheat arbitrarily

Challenge: standard cryptographic tools not feasible at this scale

our results
Our Results
  • Private computation based on secret sharing using addition only steps
    • Private addition is much simpler than multiplication
    • The main computation is in small field (32 or 64 bits) – private computation has the same cost as regular arithmetic
    • A lot of (nonlinear) algorithms can be done with addition: Regression, Classification, Bayes net, Link analysis, SVD, EM.
  • An extremely efficient ZKP that the L2 norm of user vector is bounded by L (Only O(logm) large field operations)
an efficient proof of honesty
An Efficient Proof of Honesty

The server asks for N random projections of the user’s vector, the user proves the square sum of them is small.

Projections are done in small field. The only large field operations are N encryptions and boundedness ZKP

O(log m) public key crypto operations (instead of O(m)) to prove that the L-2 norm of an m-dim vector is smaller than L.

acceptance rejection probabilities
Acceptance/rejection probabilities

(a) Linear and (b) log plots of probability of user input acceptance as a function of |d|/L for N = 50. (b) also includes probability of rejection. In each case, the steepest (jagged curve) is the single-value vector (case 3), the middle curve is Zipf vector (case 2) and the shallow curve is uniform vector (case 1)

performance
Performance

(a) Verifier and (b) prover times in seconds with N = 50, where (from top to bottom) L has 40, 20, or 10 bits. The x-axis is the vector length m.

more info
More Info
  • Code available for download, soon.
  • [email protected]
  • http://www.cs.berkeley.edu/~duan
  • Thank you!
ad