Research Project Applications Seminar SYST 798 FINAL REPORT

Research Project / Applications Seminar SYST 798 FINAL REPORT Brief Dry-Run 3 April 2008 Team: Matt Maier Tom Hare Eric Ho Brian Boynton Ali Raza Key Sponsor: Dr. Kuo-Chu Chang

Why is Smallpox a Threat ? 1972 Today 2018 • • • < 1972: Vaccination required before entering school World Health Organization declare eradication in 1977 Antibodies may decline after 10 years Population 36 years and younger = 47% USA (approx. 140 million) Militarized smallpox is only source (USA and Russia) • • Two days of life when released Variola major epidemics – 30% or higher among unvacinnated

Sponsor Information • GMU SEOR: Homeland Security and Military Transformation Lab • Dr. Kuo-Chu Chang, Professor, GMU – KChang@gmu. edu – http: //ite. gmu. edu/~kchang/ • Dr. Kathryn Blackmond Laskey, Professor, GMU – klaskey@gmu. edu – http: //ite. gmu. edu/~klaskey/

Research Conducted • Researched Technical Paperwork – 60+ articles/papers/books on biological threats, sensors, communications algorithms, disease characteristics, Chicago city characteristics, etc. – Sponsor provided over 20 technical papers in BSF research area Conducted Subject Matter Expert (SME) Interviews – Mr. Earl Zuelke, Deputy Director, Homeland Security & Emergency Management for the City of Chicago – Mr. David Hoey, Vice President, Business Development, US Genomics, DARPA BAND Biosensor Development Program – Mr. Alan Northrup, Chief Technical Officer for Sensors, Micro. Fluidic Systems, Inc. – Mr. Paul Cabellon and Ms. Alleace Gibbs, Northrop Grumman, Aerospace Systems Division, CBRNE Business Area – Dr. Paul Chew, Cornell University, Delaunay/ Voronoi Algorithm Modeling – Mr. Abbas Zaidi, CPN Modeling Earl W. Zuelke Jr. , Photo Courtesy Chicago Police Marine Unit

Research Conducted (cont. ) • Held Sponsor Meetings and Project Demos – 7 Feb, 20 Feb, 6 Mar, 20 Mar, 3 Apr • Do. D and DHS Requests for Proposal (RFPs) on Future Biosensors – Feb 2006: DARPA Biological Warfare Defense Project, $750 M+ FY 08 -FY 11 – Apr 2004: HSARPA Bioagent Autonomous Networked Detectors (BAND), Rapid Automated Biological Identification System (RABIS), $48 M 18 mo periods of performance • Researched Future Biosensor Development – Northrop Grumman Systems Corporation of Linthicum, MD – Micro. Fluidic Systems, Inc. of Pleasanton, CA – Science Applications International, Inc. of San Diego, CA – U. S. Genomics, Inc. of Woburn, MA – IQuum, Inc. of Allston, MA – Nanolytics, Inc. of Raleigh, NC – Sarnoff Corporation of Princeton, NJ – Brimrose Corporation of Baltimore, MD – Johns Hopkins University's Applied Physics Laboratory of Laurel, MD – Ionian Technologies, Inc. of Upland, CA – Goodrich Corporation of Danbury, CT – Battelle Memorial Institute of Aberdeen, MD – Physical Sciences, Inc. of Andover, MA – Research Triangle Institute of Research Triangle Park, NC

Bio. Sensor Fusion Aim • Objectives: – Minimize the time it takes to inform the public of a biological attack based on Sensor-determined dispersal of attack – Model End-to-End System for constant monitoring of urban environment • Determine optimal communications parameters and algorithm usage for Sensor Grid • Model usage of current sensor technology

Problem Statement • “Improve Urban Biological Terrorism Response” – Lack of detection and fusion today – Slow response times cost lives – False positives cost money • Biological Sensor Fusion proposes solutions for: – – – Millennium Park, Chicago. Photo Courtesy 80 s Forum Detection: Tiered sensor grid Fusion: Data Aggregation and Geo-Location Communication: Epidemic, gossip, and geographic algorithms Response: Real-time cordon mapping in changing environment Technology: State of the art in 2008 and forthcoming by 2020

Bio. Sensor Fusion Background • The Modeled Use Case – Smallpox released on the order of 10 billion organisms (~1 g) to contaminate a heavily trafficked urban area – Terrorists would spray the pathogen into the air Sears Tower, Chicago. Photo Courtesy Wikimedia Commons • System Context – Within a city the size of Chicago there is a potential for 575, 000 deaths or more – Current response plans would not allow for detection or response before 3 -4 days – Our Model will investigate employment of both current, and state of the art technology that will not be put into operation for another 10 years

Biological Sensor Fusion – System Context

Bio. Sensor Fusion Project • Bio. Sensor Team took a 3 -Pronged approach to addressing the problem of a Biological Terrorist attack in a domestic Urban environment – Architecture Products • Provides general information about system as well as providing Context for Algorithm and CPN Models as well as general information on system for client – JAVA Algorithm Model • Models the communication of the Sensor Grid upon confirmed detection of a Biological attack in an analysis of the efficacy of 5 different algorithms – Coloured Petri Nets Model Analyzes effectiveness of varying numbers and coverage areas of Sensor types, there being 3 types total

System Design Tier I: Stationary Sensors – Permanent, round-the-clock airsampling, building installed indoor and outdoor, high-regret Example: Biowatch 3 Bioagent Autonomous Networked Detector (BAND) Tier II: Mobile Ad Hoc Sensors – Deployed in emergency response vehicles (Emergency, Police, Fire, HAZMAT, etc. ) Example: General Dynamics Biological Agent Warning Sensor (BAWS) Tier III: Stationary Ad Hoc Sensors – Scattered after a threat is confirmed or incorporated in small, personal devices like cell phones – Provide low-regret tracking of dispersion Example: ICX Mesosystems Bio. Badge™ 100 Wearable Air Sampler

System Design (cont. ) • 276 sensors deployed in Chicago District 001 – 1 Operations Center and 5 additional Tier I – 120 mobile Tier II – 150 ad-hoc Tier III • Accuracy prioritized over fast detection – False alarms that shut down facilities and displace people can rival the cost of an actual outbreak (~$750 billion)

Assumptions and Constraints • • • The Bioterrorist attack occurs in District 1 of the City of Chicago, where there approximately 575, 000 people circulating as a result of the high density of attractions and tourism associated with this District A Biological terrorist attack has occurred and either Tier 1 or Tier 2 Sensors, have detected a Bioattack of smallpox virus at a minimum of 4 hours before lab analysis can occur and District Cordoning can be implemented. Evacuation plan is executed while Tier 3 sensors are additionally deployed to the specific suspected attack area. The Tier 3 sensors are deployed to further narrow down the location of the attack and isolate further zones for evacuation and cleaning The Bioterrorist attack involves the physical dissemination of ~1 g of the Smallpox organism (can fit on the head of a pin) The smallpox virus initially infects 150 people upon deployment and is deployed to only one street, limited to 1 block of potential dispersion, and will continue to infect people for 24 hrs The incubation period for smallpox is at least 7 days long; on average it takes 12 days for someone to be contagious once exposed to the virus, so infection is not being spread from person to person within the context of this system Avoiding False Positives is considered to be of prime importance: it was gleaned from our subject matter expert that a full response to a False Alarm of a Bioattack can be just as destructive as the attack itself in monetary terms

EVMS

Architecture Products • All Views – AV-1 (in development) • Operational Views – – – OV-1 OV-2 OV-3 OV-5: Node Tree & IDEF 0 OV-6 c • System Views – – – SV-1 SV-2 SV-4 SV-5 SV-6 SV-7 (in development)

CPN Model -Number of Sensor type II was cut in half -Sensor Ranges are the same -Increase of latency at early stage

CPN Model

CPN Model -Number of Sensor type II was cut in half -Sensor Ranges are the same -Number of Hops are the same

JAVA Model Analysis: Latency Conclusions: • Latency includes both re-sense time and communications time. • Latency is statistically bounded for a given range. • Latency decreases logarithmically as range increases. • Latency variance decreases with increased range. • Communications Ranges 200 m+ do not provide significant added benefit. • Sensor Ranges 150 m+ do not provide significant added benefit. • With optimum communications and sensor ranges, latency is typically 3 minutes or less

Analysis: Hop Count Conclusions: • Hop Count decreases faster than Latency (nonlinear) as range increases. This is due to the high level of disconnection in the network at low ranges. • Hop Count variance decreases as range increases. • Variance in minimum and maximum hops due to the arrival of buffered data via separate paths. • No significant improvement for ranges 250 m+ • Hop Count never decreases to less than 1 • Hop count (when optimized) is six degrees of separation or less: “Small World Communication”

Analysis: Neighbors and Coverage Conclusions: • Neighbor quantity increases exponentially with communications range, but is not affected by sensor range • Coverage increases logarithmically with sensor range, but is not affected by communications range • In both cases, variance increases as range increases • With optimal ranges, neighbors will typically be 0 -25 (largely disconnected), and coverage 30% or less

Analysis: Power Remaining Conclusions: • At low range, remaining power has a wide variance. This is due mainly to many communications hops and sensing periods, which has a large impact on power. • Low range yielded cases with still very good power conservation in the network. • Communications Ranges beyond 250 m+ have little and even sometimes a detrimental effect on power conservation. • With optimal communications, Sensor Range has a slight impact on remaining power, only at ranges <100 meters.

Analysis Conclusions • Communications Range – Optimal 250 m+ – This is feasible with a 5 watt 2. 4 GHz transmitter on ad-hoc sensors • Sensor Range – Optimal 150 m+ – This is feasible e for current biological sensors in development – Low sensor ranges provided the best geo-location accuracy • Hop Count – Optimal <6 – Hop Count and Latency are not precisely linearly related. Latency could occur while a mobile node is disconnected from the network – For speedy delivery performance, “Small World Communications” is needed • Coverage – Optimal <30% – Only impacted by Sense Range • Neighbors – Optimal 0 -25, includes disconnection – Only impacted by Communications Range • Power – [Data Needed] – Algorithm xxx provides best power conservation – Less than 10% Tier III power on average is needed when optimally configured • Fusion – A fused DHS Operations Center result is reasonable in under 5 minutes after biological agent detection. – Sense time has the most impact on overall time to respond to a biological threat.

Final Thoughts • Biological Sensor Fusion – Research into prior Biological attack/outbreak scenarios lead us to project an economic loss of $750 M and approximately 35 deaths given a status quo, lack of swift response – Our model dictates a full response within 24 -36 hours, allowing no deaths and significantly lower cost for vaccinations, cleanup, and decontamination as a result of Sensor Grid Geo. Location of threat – Our model demonstrates the fusion of data for responders to target a threat real-time, demonstrates the use of ad-hoc and mobile adhoc communications in a real-world scenario, and is a high interest research area in Do. D and DHS

Future Work • Prevention and Treatment – – Vaccination Distribution Scenarios Counter-proliferation Options Isolation and Treatment Options Emergency Response Training • Sensor Research/ Design – Advanced Technologies: UV Fluorescence, Laser-Induced fluorescence, isothermal arrays, genetic classification, electromagnetic spectroscopy, and microfluidics – Deployment Scenarios • Additional Modeling – – – Biological agent dispersal/ movement Local sensor processing and data fusion algorithms Fusion of hospital/medical practitioner data with sensor data Buffer Size, Cache, Anti-Entropy analyses Modification of model for other types of EW, ISR or CBRNE sensors Military Applications

Backup Slides

CPN Model

System Parameters • Sensor Parameters – – – – – Quantity Sensor Lat/Long Sensor Movement Range Sensitivity Specificity False Positive Rate Sense Time Sense Power Coverage • Communications Parameters – – – – Algorithm Used Range Burst Time Tx/Rx Power Reliability Data Buffering Fusion Cordon • Analysis Results – – – – Latency Delivery Rate Hop Count Coverage Remaining Power Neighbors Algorithm Type