© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Sri Elaprolu - Global Lead, Public Sector IoT, Amazon Web Services
Kristin Boorse - Sr. Product Manager, Thorn
Paul Ames - SVP, Products and Technology, SST, Inc. (ShotSpotter)
November 29, 2016
MBL 206
AWS Customers Saving Lives with
Mobile and IoT Technology
© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Kristin Boorse
November 2016
Using Machine Learning and Natural Language Processing
to Combat Human Trafficking: Going Mobile
Spotlight
THE PROBLEM
MORE THAN 200,000 NEW ADS POSTED EVERY DAY IN
THE U.S. ON SEX CLASSIFIED SITES
*Thorn Survivor Survey (2012)
OF THE UNDERAGE SEX TRAFFICKING VICTIMS SAID
THEY HAD BEEN ADVERTISED OR SOLD ONLINE*
LAW ENFORCEMENT DOESN’T HAVE ENOUGH
TIME TO NAVIGATE THE ONLINE COMMERCIAL
SEX MARKET TO FIND CHILDREN AND IDENTIFY
THEIR TRAFFICKERS.
2OO,OOO+
To provide law enforcement with intelligence and leads about suspected human trafficking networks and
individuals in order to identify victims and connect them with resources.
MISSION
Spotlight Data Processing Architecture
Elastic
Search
We
b A
PI
Elastic
Search
Escort Ad
and Forum
Websites
Amazon S3
Ad
s &
Fo
rum
s
Ph
oto
s
Amazon SNS Amazon SQS
ZData
Jetway
Core AnalyticsETL
Natural Language Processing
Phone Decoder
Language Flags (e.g.
Control)
Image Analysis
drxml
Global Analytics
Networks
Interstate Flag
Alerting
Deconstructing A Sex Trafficking Ad
Title – Tag Line
Ad copy details
Ad copy details
Ad copy details
Ad copy details
Flagging Ads
Analysis of ad copy based on risk profiles provided by law enforcement. Enables
the ability to filter based on particular targets.
MASSAGE CONTROL IMMATURE INTERSTATE CONTROL
SPOTLIGHT DATA ASSETS
150KADS PER DAY
65MADS IN SPOTLIGHT
SINCE OCTOBER 2014
400MIMAGES IN
SPOTLIGHT
THE #1 WEB-BASED DOMESTIC SEX TRAFFICKING
INVESTIGATIONS TOOL FOR LAW ENFORCEMENT
LAW ENFORCEMENT
OFFICERS USE SPOTLIGHTFEDERAL, STATE &
LOCAL AGENCIES
850+3,200+
ALL 50 STATES &
WASHINGTON D.C.
SPOTLIGHT
IMPACT
SURVEY
2,888 SURVEY RECIPIENTS
12 MONTHS
728 RESPONDENTS
25% RESPONSE RATE
VICTIMS
IDENTIFIED
6,325
HUMAN TRAFFICKING
INVESTIGATIONS
7,442
JUVENILE VICTIMS
IDENTIFIED
1,980
ADULT VICTIMS
IDENTIFIED
4,345
KIDS A DAY5+
TRAFFICKERS
IDENTIFIED
2,186
*SEPTEMBER 2016 SURVEY - LAST 12 MONTHS IMPACT
LOOK HOW FAR WE’VE COME….
* Thorn September 2016 user survey [25% response rate]
60%
DAILY SPOTLIGHT USERS
REPORT A 60% TIME SAVINGS IN
HUMAN TRAFFICKING
INVESTIGATIONS
BEFORE SPOTLIGHT AFTER SPOTLIGHT
””
Spotlight is a force multiplier at every stage of the operation. We can
use this tool to gather intelligence and plot trends prior to conducting
an operation, during the operation we can allocate resources effectively
and tighten our investigative vision to have the greatest immediate
impact… Federal Agent, Hawaii
””
A juvenile runaway was kidnapped from Dallas and taken to Houston
where she was forced to stay in motels under armed guard. Her picture
was put on backpage.com along with a kidnapper's phone number. We
identified one of the "johns" who raped the juvenile and identified all of the
phone numbers that he called with a search warrant. We ran all of the
phone numbers through spotlight until we found the backpage.com ads
that listed the juvenile victim. We also gained our only lead on who the
kidnappers were through the backpage.com account.
Investigator Bocks, Houston PD
Thank you!
Facebook.com/wearethorn
Twitter: @thorn
Kristin Boorse
Senior Product Manager
480.600.1468
© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Paul Ames
November 2016
SST ShotSpotterPublic Cloud Migration
ShotSpotter cities experienced an
average 35% DECREASE
in gunfire incident volume
in the first two years of use
35%DECREASEgunfire incident
volume
2015 National Gunfire Index report
Raw
Incidents
Raw
Incidents
Raw
Incidents
Sensor
Location
Machine Classification
Human Classification
Incident Publisher
Email SMSAdminReviewer Patrol InvestigatorNotification
API
Security
Auth
entication
Unstr
uctu
red D
ata
Sto
rage
Monitoring
Sensor Sensor Sensor Sensor Sensor
Architectural starting point
Private cloud deployment
Mobile Carrier
Machine-2-Machine
Private Network
Public
InternetSST Private Cloud West Coast (Colo)
SST Private Cloud East Coast (Colo)
SMS
Admin
Reviewer
Patrol
Investigator
Notification
API
City A
Sensor Array
Campus
Sensor Array
City B
Sensor Array
City C
Sensor Array
Commodity x86, VMWare
Hybrid cloud deployment exploration
Mobile Carrier
Machine-2-Machine
Private Network
Public
Internet
Public Cloud - Lift & Shift
SST Private Cloud West Coast (Colo)
SST Private Cloud East Coast (Colo)
SMS
Admin
Reviewer
Patrol
Investigator
Notification
API
City A
Sensor Array
Campus
Sensor Array
City B
Sensor Array
City C
Sensor Array
Commodity x86, VMWare
Phase 1 implementation
Mobile Carrier
Machine-2-Machine
Private Network
Public
Internet
AWS Public Cloud
SST Private Cloud West Coast (Colo)
SST Private Cloud East Coast (Colo)
SMS
Admin
Reviewer
Patrol
Investigator
Notification
API
City A
Sensor Array
Campus
Sensor Array
City B
Sensor Array
City C
Sensor Array
Commodity x86, VMWare
Amazon EC2/CentOS,
Amazon ECS/Docker, ALB,
Amazon VPC, Amazon Route 53
Microservices: Node.js, MongoDB,
Reddis, RabbitMQ, APNS & GCM
Web
Mobile