18
PACT: Private Automated Contact Tracing Presented at the IEEE EMBS 2020 Annual Conference Marc Zissman, PhD MIT Lincoln Laboratory 24 July 2020 Sponsors: DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited. This material is based upon work supported by the Defense Advanced Research Projects Agency under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Defense Advanced Research Projects Agency. © 2020 Massachusetts Institute of Technology. Delivered to the U.S. Government with Unlimited Rights, as defined in DFARS Part 252.227-7013 or 7014 (Feb 2014). Notwithstanding any copyright notice, U.S. Government rights in this work are defined by DFARS 252.227-7013 or DFARS 252.227-7014 as detailed above. Use of this work other than as specifically authorized by the U.S. Government may violate any copyrights that exist in this work.

PACT: Private Automated Contact Tracing · 2020. 7. 24. · • PACT seeks to advance the state-of-the-art in private automated contact tracing solutions • Significant opportunities

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

  • PACT: Private Automated Contact Tracing

    Presented at the IEEE EMBS 2020 Annual Conference

    Marc Zissman, PhDMIT Lincoln Laboratory

    24 July 2020

    Sponsors:

    DISTRIBUTION STATEMENT A. Approved for public release. Distribution is unlimited. This material is based upon work supported by the Defense Advanced Research Projects Agency under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the Defense Advanced Research Projects Agency. © 2020 Massachusetts Institute of Technology. Delivered to the U.S. Government with Unlimited

    Rights, as defined in DFARS Part 252.227-7013 or 7014 (Feb 2014). Notwithstanding any copyright notice, U.S. Government rights in this work are defined by DFARS 252.227-7013 or DFARS 252.227-7014 as detailed above. Use of this work other than as specifically authorized by the U.S. Government may violate any copyrights that exist in this work.

  • Page 2MAZ 07/24/2020

    COVID-19 Infection Progression

    Exposure

    10 days 20 days5 days 15 days 25 days0

    InfectedContact

    Symptom Onset

    IndexExposure

    IndexCase Infectious

    Recovery

    Recovery

    Infectious

    Incubation

  • Page 3MAZ 07/24/2020

    COVID-19 Infection Progression

    Exposure

    10 days 20 days5 days 15 days 25 days0

    InfectedContact

    Symptom Onset

    IndexCase

    https://www.cdc.gov/coronavirus/2019-ncov/hcp/clinical-guidance-management-patients.htmlThe Incubation Period of Coronavirus Disease 2019 (COVID-19) From Publicly Reported Confirmed Cases: Estimation and Applicationhttps://wwwnc.cdc.gov/eid/article/26/7/20-0282_article https://www.harvardmagazine.com/2020/05/r-noughthttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC7054855/ ; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7121626/

    Infectious

    Recovery

    Recovery

    Goal: Find this person before they might infect others• Identify “contacts” that could have infected this person (reverse)• Identify “contacts” that this person could infect (forward)

    IndexExposure

    Incubation

    Infectious

    https://www.cdc.gov/coronavirus/2019-ncov/hcp/clinical-guidance-management-patients.htmlhttps://www.acpjournals.org/doi/abs/10.7326/M20-0504https://wwwnc.cdc.gov/eid/article/26/7/20-0282_articlehttps://www.harvardmagazine.com/2020/05/r-noughthttps://www.ncbi.nlm.nih.gov/pmc/articles/PMC7054855/https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7121626/

  • Page 4MAZ 07/24/2020

    • Traditional uses:– Tuberculosis (2.9 cases / 100k people,

    > 900 cases / 100k for COVID-19 Ϯ )– Smallpox– Sexually transmitted diseases

    Contact Tracing

    Contact tracing is an epidemiological technique used to identify people who have had “contact” with an infected person

    Contact tracing can help inform public health interventions to slow virus transmission

    Infected

    Uninfected

    Time

    Transmission

    Ϯ https://www.cdc.gov/tb/statistics/indicators/2018/incidence.htm for Massachusettshttps://www.cdc.gov/mmwr/volumes/69/wr/mm6915e4.htm

    https://www.cdc.gov/tb/statistics/indicators/2018/incidence.htmhttps://www.cdc.gov/mmwr/volumes/69/wr/mm6915e4.htm

  • Page 5MAZ 07/24/2020

    • Prior to COVID-19, contact tracing was primarily a manual process• Primarily used for diseases with longer temporal characteristics

    Contact Tracing Tools

    Public Health Contact Tracing ToolsChallenges• Index case has to remember who they

    were in contact with, where they were

    • Labor intensive and time consuming

    • Increased risk of data errors

    • Difficult to apply analytics

    • Does not scale to need

    • Need to know identifying information for contacts

    Advanced contact tracing tools are urgently needed to handle COVID-19

  • Page 6MAZ 07/24/2020

    Animation Video

    Animation Video: https://www.youtube.com/watch?v=yuXzAh4slNw

    https://www.youtube.com/watch?v=yuXzAh4slNw

  • Page 7MAZ 07/24/2020

    Likely Late Summer Scenarioin a U.S. State

    Approx 6 feet

    ChirpLog

    Database ofAnonymized Contacts

    BluetoothChirp

    RecordsInfectionCertification

    ChirpLog

    BluetoothChirp

    Contact Distance

    IndexCase

    Contact IndexCase

    Contacts

    ManualContacts Records

    Contact Detection

    1 COVID-19 Test Confirmed Positive

    2 Exposure Query & Notification

    3 Contact Actions

    4

    State Dept of HealthVirtual Community Tracing Center

    ManualNotification

    AutomaticQuery &

    Notification

    Test

    Symptom check

    SelfQuarantine

    Contacts

    iPhone orAndroid

    • Updated Apple & Google OS w/BT contact tracing built-in

    • State-acquired and A|G-approved app for user interface

    1

    • State-acquiredsynchronized database

    2

    • State-acquired appwill provide tailoredinstruction to contacts

    3

  • Page 8MAZ 07/24/2020

    “Too Close for Too Long” – TC4TL

    Range

    Expo

    sure

    Tim

    e

    0m

    10min

    Hug

    30min

    1hr

    8hrs

    1m 2m 3m 4m

    1min

    SameClassroom

    Subway,One Stop,Same Car

    HandshakeNo Alert

    • Public health authorities define alert region• Engineers try to implement it accuratelyALERT!

    Side-by-Side@ Coffee Shop

    SameTable @School

    NotionalAlert Region

  • Page 9MAZ 07/24/2020

    Assessing TC4TL Systems

    Probability of False Positive(Leads to applying effort & treatment unnecessarily to a healthy person)

    Prob

    abili

    ty o

    f Fal

    se N

    egat

    ive

    (a.k

    .a. M

    iss)

    (Lea

    ds to

    with

    hold

    ing

    effo

    rt &

    trea

    tmen

    t fro

    m a

    sic

    k pe

    rson

    )

    0.01% 99%0.1% 1% 10% 50%

    “Flip a Coin”

    0.01

    %99

    %0.

    1%1%

    10%

    50%

    System A

    System BEverybodyStay Home

    Do Nothing

    Better SystemPerformance

    *A. Martin, A., G. Doddington, T. Kamm, M. Ordowski, and M. Przybocki. "The DET Curve in Assessment of Detection Task Performance",Proc. Eurospeech '97, Rhodes, Greece, September 1997, Vol. 4, pp. 1895-1898.

    IdealPerformance

    • Want to give public health authorities freedom to pick from a range of operating points

    • If we can estimate Pr(pos), Pr(neg), Cost(FN), Cost(FP), we can find the lowest-cost operating point on the DET* curve

    Decision Error Tradeoff (DET*) Curve

  • Page 10MAZ 07/24/2020

    Initial Bluetooth Data Collection

    Generate “Near” pdfs Generate “far” pdfs

    • Empirically measure RSSI vs range for many conditions

    – Phone orientation– Location of phone– Phone model – Multipath environment

    Probability Density Function of RSSI vs RangeRotation through 3600

    RSS

    I (dB

    m)

    Distance (ft)

    Prob

    abili

    ty

    RSSI Variability

    Phone conditions lead to 20dB (100x) variation in RSSI at fixed distance

  • Page 11MAZ 07/24/2020

    Preliminary TC4TL Performance**

    Decision Error Tradeoff (DET) Curve* • 1 of 6, Agnostic: 30 min duration,phone wakes up 6 times, one attenuation per wake up period, no info re phone carriage

    • M of 24, Agnostic: 30 min duration,phone wakes up 6 times, 4 samplesper wake up period, no info re phone carriage

    • 1 of 6, Cognitive: Same as 1 of 6 above except threshold specific to perfect phone carriage information

    • M of 24, Cognitive: Same as M of 24 aboveexcept threshold specific to perfect phone carriage information

    • Data: “Range-Angle” data set. Collected at distances of 3 to 15 feet at every 45-degree angle for a variety of conditions (e.g. phone in pocket, hand, purse, bag, etc.)

    Better signal processing with additional metadata sent with chirp(no additional power, storage) can reduce EER from ~40% to ~15% EER

    *A. Martin, A., G. Doddington, T. Kamm, M. Ordowski, and M. Przybocki. "The DET Curve in Assessment of Detection Task Performance",Proc. Eurospeech '97, Rhodes, Greece, September 1997, Vol. 4, pp. 1895-1898.

    ** G. F. Hatke et al. “Using BLE Signal Strength Estimation to Facilitate Contact Tracing for COVID-19” https://arxiv.org/abs/2006.15711

    Equal ErrorRate (EER) Line(PFA = Pmiss)

    https://arxiv.org/abs/2006.15711

  • Page 12MAZ 07/24/2020

    Advanced Proximity Testing

    Mannequins on

    TurtleBots

    Robotic Testing in the ASDF MIT LL @ Waypoint Robotics

    Bluetooth

    Scanning Apps

    • LL custom• 3rd party

    ASDF

    ASDF = MIT LL Autonomous System Development Facility (Group 76)

    Animation Video: "Robot Dance" https://youtu.be/1F6TEvpy5g0

    https://youtu.be/1F6TEvpy5g0

  • Page 13MAZ 07/24/2020

    PACT Public Data Repositoryhttps://mitll.github.io/PACT/

    https://mitll.github.io/PACT/

  • Page 14MAZ 07/24/2020

    Evaluations: NIST TC4TL Challengehttps://www.nist.gov/itl/iad/mig/nist-tc4tl-challenge

    • Explore promising new ideas in TC4TL detection using BLE signal• Support the development of advanced technologies incorporating

    these ideas• Measure and calibrate the performance of the state-of-the-art

    TC4TL detectors

  • Page 15MAZ 07/24/2020

    User Interfaces

    Epidemiology DashboardsMobile Phone App

    • For individual citizens

    • Alerts user of potential exposure

    • Turns exposure tracing on and off

    • For public health professionals

    • Helps contact tracers manage cases

    • Provides summary epidemiological metrics

  • Page 16MAZ 07/24/2020

    PACT Accomplishments

    ImPACT 2020 Conference

    Exposure notification system for iOS and Android

    • Welcome remarks by MA Gov Baker

    • Technical exchange of ideas

    • 500+ global participants

    • A|G ENS consistent with the PACT protocol

    • Providing technical advice to Apple and Google– “Too close for too long” detector design– Laboratory RF measurements and analysis

    • Weekly meetings with the A|G technical teams

    • Trusted technical advisor to US CDC

    • Trusted technical advisor to Massachusetts and Pennsylvania

  • Page 17MAZ 07/24/2020

    Advanced Concepts

    Improved Inter-Device Ranging “Wearables” and Tokens

    Apple | Google Subsystem Compatibility Trustless Servers and Data Stores

    UltrasoundUltra wideband(G39’s LLDART) Smart watches

    RF tokens & bands

    Custom Hardware

    Key Exchange

    Bluetooth Chirps

    LLDART = Low-cost Localization using Distributed Adaptable Response Transponders

  • Page 18MAZ 07/24/2020

    • Contact tracing combined with public health action, such as testing and quarantining, will help reduce virus spread

    • Automated contact tracing can supplement manual efforts– Automation of contact analysis– Automatic detection of high-risk exposure events

    • PACT seeks to advance the state-of-the-art in private automated contact tracing solutions

    • Significant opportunities for future technical innovation exist

    Summary

    PACT serves as trusted technical advisor to federal, state and local public health authorities (PHAs) and as a convening center for collecting and sharing data and best

    practices for private automated contact tracing