DEPLOYING MPC DE FOR SOCIAL GOOD FO Lucy Q Lu Qin in Re - - PowerPoint PPT Presentation

Lu Lucy Q Qin in

Re Resear arch Soft ftwar are Engineer | | Soft ftwar are & Applicat ation Innovat ation Lab ab | | Boston Un University Andr Andrei i La Lapets, , Frederick Jan ansen, , Ki Kinan an Ba Bab, Peter Floc Flockha hart, Raw Rawan ane Is Issa, , Ma Mayank Va Varia, , Az Azer Be Bestavros

• s

DE DEPLOYING MPC FO FOR SOCIAL GOOD

MPC FOR SOCIAL GOOD

Pay Equity Combatting Opioid Addiction Medical Data Sharing Economic Inclusion Social Science Research Sharing Detecting Serial Perpetrators of Sexual Misconduct

MPC FOR SOCIAL GOOD

3

Pay Equity Combatting Opioid Addiction Medical Data Sharing Economic Inclusion Social Science Research Sharing Detecting Serial Perpetrators of Sexual Misconduct

SECURE MULTI-PARTY COMPUTATION (MPC)

f(s1, s2, s3) = z

pu publ blic

• u
• utput

pri rivate inputs

CLOSING THE WAGE GAP IN BOSTON

1. 1. Un Understan anding the root ca causes of th the wage gap 2. 2. Cl Closi sing th the gap 3. 3. Ev Evaluati ting su succe ccess ss

Over 200 Businesses

Business A Trusted 3rd Party Results! Business B Salary Data Salary Data

ORIGINALLY PROPOSED WORKFLOW

Business A Trusted 3rd Party Results! Business B Salary Data Salary Data

ORIGINALLY PROPOSED WORKFLOW

BWWC IT Personnel Lawyer HR/Diversity Personnel

ROLES & CONCERNS

ROLES & CONCERNS

Lawyer Lia Liabilit ility BWWC IT Personnel HR/Diversity Personnel

ROLES & CONCERNS

BWWC Pa Participation Lawyer Lia Liabilit ility IT Personnel HR/Diversity Personnel

ROLES & CONCERNS

BWWC Pa Participation Lawyer Lia Liabilit ility HR/Diversity Personnel Us Usab ability IT Personnel

ROLES & CONCERNS

BWWC Pa Participation IT Personnel Au Audit ditabilit ility Lawyer Lia Liabilit ility HR/Diversity Personnel Us Usab ability

BU S BU Server (w (web se server/database se)

Analyst can never access this data

Con Contribu butor

• r B

Con Contribu butor

• r A

random mask A actual data A

+

masked data A

= An Analyst st at BW BWWC (client running web browser)

random mask B

+ =

actual data B

Public-key Encrypted Storage

• nly Analyst has key;

no one else (including BU) can read the content of this data

masked data A masked data B masked aggregate data

+ =

random mask A random mask B random mask A random mask B

+ =

aggregate mask masked aggregate data

_ =

actual aggregate data masked data B

MPC SOLUTION

100talent.org

15

ERROR CHECKING

USER EXPERIENCE & WEB ANALYTICS

USER EXPERIENCE & WEB ANALYTICS

2016 2017 total # of employers 69 114 # employees (1000s) 113 167 % of workforce 11 16 total annual earnings $11b $15b

MPC ENABLED FOR DATA ANALYSIS

LONGITUDINAL DATA

IMPROVED DATA GRANULARITY

FROM DATA TO IMPACT

SUPPORTING LOCAL MINORITY-OWNED BUSINESSES

DETECTING SERIAL PERPETRATORS OF SEXUAL MISCONDUCT

6x 15% 3x

more likely to seek out me medical and emo motional su support rt services aft after us using Ca Callisto sto

• f

f su surv rvivors rs were re matched wi with another survi vivo vor wh who re reporte rted th the sa same per perpet petrator more likely to report th thro rough a sch school’s s Ca Callisto sto we webs bsite than to th their r sch school or r th the police ce

Identifying information about a survivor and the accused can only be decrypted by a lawyer when at least 2 users name the same perpetrator

Bob Carlos Alice Accused: Mallory Accused: Mallory Accused: Eve Alice Mallory Bob Mallory Carlos Eve

DETECTING SERIAL PERPETRATORS OF SEXUAL MISCONDUCT

where (x, y) is a share

y = ax + k

ke key re repre rese sented as as y-in inter ercep ept sl slop

• pe

DETECTING SERIAL PERPETRATORS OF SEXUAL MISCONDUCT

ha hashe hed d us user er id id

hashed user id (x) share value (y) Carlos Alice Bob

/multiparty/umbral /multiparty/oprf

27

DEPLOYING MPC FOR SOCIAL GOOD

@lc lcyqn @bu bu_compu putin ing @hi hics csail /m /multiparty /hi hics csail mu multiparty.org sai sail.bu.edu

THANK YOU

29