SLIDE 1
2/23
Green Security How Can AI Help in Protecting Forests, Fish and - - PowerPoint PPT Presentation
Green Security How Can AI Help in Protecting Forests, Fish and - - PowerPoint PPT Presentation
Green Security How Can AI Help in Protecting Forests, Fish and Wildlife MILIND TAMBE Helen N. & Emmett H. Jones Professor in Engineering University of Southern California WHAT MIGHT WE LOSE? Murchison Falls National Park, Uganda 2/23
SLIDE 2
SLIDE 3
3/23
Murchison Falls National Park, Uganda
WHAT MIGHT WE LOSE?
SLIDE 4
PAWS: PROTECTION ASSISTANT for WILDLIFE SECURITY
Massive forests (1000 sq miles) to protect, limited security resources:
- How to Efficiently Patrol/Protect forests with limited resources
- PAWS patrols: Exploit past poaching data, avoid predictability
Patrol boat in Bangladesh at Global Tiger Conference, 2014 Patrol with Rangers, Indonesia Trip with WWF, 2015
4/23
SLIDE 5
5/23
Game Theory
AI-based DECISION AIDS TO ASSIST IN SECURITY
2007
Airports
SLIDE 6
6/23
AI-based DECISION AIDS TO ASSIST IN SECURITY
SLIDE 7
7/23
Game Theory
2007
Airports
- 1, 1
0, 0 1, -1 1, -1
- 1, 1
0, 0
Player B Player B Player A Player A
Paper Rock Scissors Paper Rock Scissors
0, 0 1,-1
- 1,1
AI-based DECISION AIDS TO ASSIST IN SECURITY
SLIDE 8
8/23
Game Theory
2007
Airports Canine patrol at LAX (ARMOR)
AI-based DECISION AIDS TO ASSIST IN SECURITY
SLIDE 9
9/23
2007 2011
Airports Ports Air Marshals
2009
Game Theory
AI-based DECISION AIDS TO ASSIST IN SECURITY
SLIDE 10
PROTECT: FERRY PROTECTION DEPLOYED [2013-]
10/23
SLIDE 11
11/23
2007 2011 2013
Airports Ports Trains Air Marshals
2009
Game Theory
AI-based DECISION AIDS TO ASSIST IN SECURITY
SLIDE 12
GLOBAL PRESENCE OF SECURITY USING GAME THEORY
12/23
SLIDE 13
13/23
- Game theory vs Previous Method
Ticketless Travelers Caught Arrests at LAX checkpoints
5 10 15 20
#Captures per 30 min #Warnings per 30 min #Violations per 30 min
Game Theory Previous Method
20 40 60 80 100 Miscellaneous Drugs Firearm Violations
SOME RESULTS OF GAME THEORY for SECURITY
Game Theory in the Field
SLIDE 14
14/23
GAME THEORY FOR PATROLS [2013]
Congressional Subcommittee Hearing
SLIDE 15
15/23
Learn from crime data Game Theory calculate randomized patrols
Patrollers execute patrols Poachers attack targets
Predicting Poaching from Past Crime Data
PAWS: APPLYING AI FOR PROTECTING WILDLIFE
Game Theory + Poacher Behavior Prediction
SLIDE 16
16/23
How likely is an attack on a grid Square
Ranger patrol frequency Animal density Distance to rivers / roads Area habitat Area slope
…
Queen Elizabeth National Park, Uganda
12 years of patrols, 125000 observations
POACHER BEHAVIOR PREDICTION
SLIDE 17
17/23
Game Theory + Poacher Behavior Prediction
12 years of patrols, 125000 observations
Dry Season (June-August 2008)
PAWS INITIAL SYSTEM [2016]
SLIDE 18
18/23
Uganda
Andrew Lemieux
Panthera Malaysia
PAWS PATROLS IN THE FIELD [2016]
Trials in Uganda and Malaysia
Important Lesson: Geography!
SLIDE 19
19/23
PAWS: PROTECTION ASSISTANT FOR WILDLIFE SECURITY [2016]
Game Theory + Poacher Behavior Prediction + Forest Street Map
SLIDE 20
PAWS: PRELIMINARY EVALUATION
0.57 0.86
Human Activity Sign/km
Previous Patrol PAWS Patrol
20/23
SLIDE 21
PAWS COLLABORATIONS
21/23
SLIDE 22
22/23
FISHERY PROTECTION FOREST PROTECTION
AI DECISION AIDS for PROTECTING FORESTS, FISHERIES, RIVERS
Protecting Forests, Fish, Rivers
RIVER POLLUTION PREVENTION
SLIDE 23
AI and GAME THEORY WORLDWIDE FOR SOCIAL GOOD
23/23
Thank you to sponsors:
SLIDE 24
THANK YOU
tambe@usc.edu http://teamcore.usc.edu/security
24/64
SLIDE 25
25/64
Lab Evaluation Simulated adversary Human subject adversaries
Security Games superior vs Human Schedulers/”simple random”
EVALUATING DEPLOYED SECURITY SYSTEMS NOT EASY How Well Optimized Use of Limited Security Resources?
Field Evaluation: Patrol quality Unpredictable? Cover? Compare real schedule Scheduling competition Expert evaluation Field Evaluation: Tests against adversaries “Mock attackers” Capture rates of real adversaries
SLIDE 26
26/64
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Count Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Count Base Patrol Area
Patrols Before PROTECT: Boston Patrols After PROTECT: Boston
PROTECT (Coast Guard)
FIELD EVALUATION OF SCHEDULE QUALITY
Improved Patrol Unpredictability & Coverage for Less Effort
: 350% increase in defender expected utility
SLIDE 27
27/64