Prof. Gautam Dasgupta Columbia University Security Engineering -- An Introduction to the LiveDesign...

Prof. Gautam DasguptaColumbia University

Security Engineering-- An Introduction to the

LiveDesign Paradigm

byGautam Dasgupta, Professor, Civil EngineeringColumbia University, New York, NY 10027, USA

Tel:212-854-3102;fax:212-854-6267;email:[email protected]

url:www.columbia.edu/~gd18

Keywords: Bayesian updating, extreme value statistics, security and sensors

Date modified: August 8, 2004


Design vs. LiveDesign

Conventional design/construction governed by:

1. mechanical stress

2. cost considerations

LiveDesign

a. geared towards mitigating disasters

b. detects imminent threats and breach of security


IT based LiveDesign

Information Technology (IT) has empowered engineers

to develop LiveDesign prototypes that

I) alert civic authorities confronted

with manmade disasters

II) suggest optimal evacuation/rescue

along the most favorable routes


Statistical decision in LiveDesign

IT-based optimization/decision process

1. high speed

2. large scale

3. distributed computation

4. impliments extreme-value statistics

Bayesian updating is almost instantaneous

Hence the paradigm is termed LiveDesign.


1. Background work from 2001(a) engineers(b) architects(c) computer scientists(d) lawyers

2. LiveDesign core methodology3. Randomly selected office building models4. Seamless information augmentation -- open

environment5. Nonrestrictive extension for all civil infrastructures

(a) watersheds(b) energy systems

6. Versatility of LiveDesign

LiveDesign Research


Three principal parts of LiveDesign

1. Data acquisition

2. Statistical combination of characteristic indicators

to construct threat scenarios

3. Display/broadcast virtual reality-based

smart signs of optimal paths:

a. evacuation

b. rescue


Data acquisition combines:

1. Available structural stress analysis

and response history

2. Building codes

3. Legal requirements

4. On-line sensor readings

Quantitative and qualitativeData in LiveDesign


Threat scenarios constructed by

statistical combination:

1. Principal component analysis

assesses the most fatal state

2. Interval arithmetic depicts

confidence bounds

3. fuzzy logic constructs

spatio-temporal threat patterns

LiveDesign


Functions of LiveDesign IT engines

1. carry out dynamic searches on the LiveDesign database

1-a. signal fatal events from extreme value statistics

1-b. check tight bounds on the confidence interval to predict credible threats

2. sort out anomalies such as noise and spreads on the confidence bounds

2-a. continuously fine-tune a self-learning process

2-b. eliminate false alarms as statistical outliers

3. discover yet to be identified ominous impending events


LiveDesign


LiveDesign Bayesian Network1. Security costs depend nonlinearly on the associated payoff

2. All indices can be assessed only in the probabilistic sense

3. Bayesian network is an elegant representation:(a) operates on joint probabilities(b) uses graphical communication language(c) is based on the semantics of causality

4. Modeling procedure is symbolic has been partially written in Mathematica5. Current research contains Modelica:— graph theory constructs implement qualitative features


Elements of the network

1. The network formulation starts with elementary subsystems

2. Modelica can create complex systems using the formal specifications of

the links in the LiveDesign database

3. Benchmark examples have been verified using Mathematica:


Large number of variables

1. Eigenvalues with Interval Arithmetic

— principal component analysis

2. Yield the threat index from a composite scenario

3. An order of a thousand statistical security variables

4. The generic form:

integral equations

yields the fixed point scheme with the kernel .

This has already been conveniently programmed in Mathematica.

∫= ηηη dyfyxfyxf )|(),|()|(

∫= ηηη dyfxfyxf )|()|()|(

∫ ∫== dxxfxyfyfdyyfyxfxf )()|()(;)()|()(

∫= ')'()',()( dxxfxxhxf

∫= dyxyfyxfxxh )'|()|()',(


Extreme Value Statistics

Table 1: Three forms of distributions of extremes


Chemistry experiments to bridge theory and reality based learning and provide the bases for bio-sensors:

1. There may be more than a single answer to a question or problem

2. Several outcomes may be possible with a varying degree of predictability

3. The fuzzy logic concepts resolve such issues


LiveDesign deep domain modeling formalism

1. Architects create the LiveBluePrint2. A collection of its various parts, and then assembled.3. Modern buildings are more complicated:

(a) there are additional subsystems(b) such as power, water, elevators(c) and above all a security infrastructure


Intelligent Infrastructureby LiveDesign

1. Smart structures respond to their surroundings

(a) Safety

(b) Thermo-mechanical inputs

2. Intelligent infrastructures

(a) Security

(b) Safety concerns as subsets

(c) Societal policy inputs

3. Information based technology

(a) Integration of quantitative and qualitative data

(b) Computer understandable . representation/communications


The LiveDesign Team

1. Universities(a) Columbia University(b) MIT(c) Carnegie Mellon University(d) University of Maine

2. Industries(a) OKG Constructions, NYC(b) Contour Graphics Architects, NYC(c) Serby Attorneys, NYC

3. Foreign collaborators(a) Kagawa University, Japan

(b) Laboratoire d’Optique, Besancon, France

Prof. Gautam Dasgupta Columbia University Security Engineering -- An Introduction to the LiveDesign...

Documents

Transcript of Prof. Gautam Dasgupta Columbia University Security Engineering -- An Introduction to the LiveDesign...