Dr. Praveen Paruchuri: His ARMOR Security System Uses Unpredictability to Help Secure Police Operations at LAX

November 7, 2008

Dr. Praveen Paruchuri is currently working as a Research Scientist at Intelligent Automation Inc. (IAI), Rockville, MD. He earned his PhD in Multiagent Systems from USC’s computer science department in August 2007. Praveen's research interests include Applied AI, Agents/Multiagent Systems, Decision/Game Theoretic Reasoning, Safety/Security Issues and Linear/Nonlinear/Mixed Integer programming. He was presented the Best Paper Award at the SASEMAS International workshop for developing security algorithms using the MDP/Dec-POMDP frameworks. More recently, his game-theoretic algorithm named DOBSS became the heart of the ARMOR system that is currently being deployed at the LA International Airport (LAX). This work has been described in numerous radio reports, Channel 4 KNBC news and news media such as LA Times, Newsweek, International Herald Tribute, Times of India, Lenta.Ru, New Scientist etc. More details about Praveen are available on his website.

We met with Dr. Paruchuri and asked him a few questions:

Help us understand what you are up to:
My doctoral thesis was on the topic of "Security of Multiagent Systems" and it led to the development of the ARMOR (Assistant for Randomized Monitoring over Routes) security scheduling system. Let’s consider the problem of assigning security checkpoints on various roadways entering airport for monitoring purposes. Given the limited security personnel available, airports cannot afford to have checkpoints on all roads at all times. Furthermore, adversaries can monitor the checkpoints regularly and learn weaknesses/patterns. ARMOR takes into account various factors like the number of possible checkpoints, their operation times, traffic patterns, estimated target priority for adversary, cost for adversary to get caught etc. and calculates the optimal randomized monitoring solution.

More specifically, ARMOR casts the security problems like patrolling/monitoring into a Bayesian Stackelberg game. In a Stackelberg game, a leader commits to a strategy first, and then a follower selfishly optimizes its reward, considering the action chosen by the leader. A Bayesian Stackelberg game is a Stackelberg game where the leader may face multiple follower types (i.e. each adversary can appear with some probability). In most security domains, while the police/canine units have to commit to a security policy, adversaries can observe for a long time and exploit the policy committed to. I mapped this key observation to a Bayesian Stackelberg game and developed the fastest solver for it till date named DOBSS (Decomposed Optimal Bayesian Stackelberg Solver) which became heart of the Armor system. 

What drives you to continue pursuing this area of study?
The field of AI in general provides many opportunities to see my ideas work in real life in a shorter timeframe rather than being limited to papers or computer labs.

Describe how your work might impact people’s lives, now and in the future.
The Armor system successfully completed a six month trial deployment in Feb’08 at the Los Angeles International Airport (LAX) for randomizing security checkpoints and scheduling canine patrols. During the review meet LAX officials noted that the software has become the new gold standard for security checkpoints and canine patrols. This application got extensive press coverage and has been featured in various national media outlets like Newsweek, LA Times, The International Herald Tribute, Government Computer News, Homeland Security Today etc. and International newspapers like Times of India (India), Lenta.Ru (Russia) etc. The story also appeared on numerous radio reports and channel 4 KNBC News.

Future Potential Applications:  Given the numerous opportunities that opened up in a short duration since its development in 2007, I envision my software assistants helping security officials in making critical decisions for enhancing security during general policing, border patrolling and at locations of economic or political importance like airports, dams, museums, stadiums, important buildings etc.

How did you come up with the idea?
Out of necessity to come up with something strong to finish my PhD J. Jokes apart, halfway through my PhD I had a good theoretical framework without a real application. Plenty of discussions later with Milind and others, I zeroed in on security as a potential application.

Has anyone ever doubted that your idea could work?
Yes. And surprisingly it wasn’t the LAX officials or USC Campus Police. They were very appreciative of the idea in the initial talks itself.

The people who doubted most were reviewers at AI conferences who rejected the idea initially on grounds that developing randomization algorithms for security isn’t a strong enough motivation (Maybe the AI Winter effect !!!).

What is the next step in the innovation process for you?
Making millions ;)

What is the one innovation you can’t live without?
I guess the standard answer of geeks: The Internet

What do you wish you would have invented? 
Computers

Any tips for aspiring innovators?
Chase your ideas. The process gives lots of happiness irrespective of the results.

Email or phone? 
Email

What is the most fun you’ve ever had?
Tough call.

Three favorite things about LA:
Weather
Beaches
Proximity to Vegas ;)