Modeling, Simulation and Analysis of

Complex Networked Systems

A Program Plan
May 2009

Authors:
James M. Brase
David L. Brown
Lawrence Livermore National Laboratory
May 2009
Front and back covers: Diagrams of the Abstract Syntax Tree (AST) internal
representation of a simple C++ program. The firework burst patterns are
namespaces and different parts of the internal classes, functions and code within
functions. The front cover shows the AST for source code, while the back cover
shows a similar AST used to represent the structure of binary executables for
analysis.
Source: Dan Quinlan, Center for Applied Scientific Computing, Lawrence Livermore
National Laboratory

Disclaimer
This document was prepared as an account of work sponsored by an agency of the United States government. Neither
the United States government nor Lawrence Livermore National Security, LLC, nor any of their employees makes any
warranty, expressed or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or
usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe
privately owned rights. Reference herein to any specific commercial product, process, or service by trade name,
trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or
favoring by the United States government or Lawrence Livermore National Security, LLC. The views and opinions of
authors expressed herein do not necessarily state or reflect those of the United States government or Lawrence
Livermore National Security, LLC, and shall not be used for advertising or product endorsement purposes.
LLNL-TR-412733

This work performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory
under Contract DE-AC52-07NA27344.

Executive Summary
Many complex systems of importance to the U.S. Department of Energy consist of
networks of discrete components. Examples are cyber networks, such as the
internet and local area networks over which nearly all DOE scientific, technical and
administrative data must travel, the electric power grid, social networks whose
behavior can drive energy demand, and biological networks such as genetic
regulatory networks and metabolic networks. In spite of the importance of these
complex networked systems to all aspects of DOEs operations, the scientific basis
for understanding these systems lags seriously behind the strong foundations that
exist for the physically-based systems usually associated with DOE research
programs that focus on such areas as climate modeling, fusion energy, high-energy
and nuclear physics, nano-science, combustion, and astrophysics.
DOE has a clear opportunity to develop a similarly strong scientific basis for
understanding the structure and dynamics of networked systems by supporting a
strong basic research program in this area. Such knowledge will provide a broad
basis for, e.g., understanding and quantifying the efficacy of new security
approaches for computer networks, improving the design of computer or
communication networks to be more robust against failures or attacks, detecting
potential catastrophic failure on the power grid and preventing or mitigating its
effects, understanding how populations will respond to the availability of new
energy sources or changes in energy policy, and detecting subtle vulnerabilities in
large software systems to intentional attack.
This white paper outlines plans for an aggressive new research program designed to
accelerate the advancement of the scientific basis for complex networked systems of
importance to the DOE. It will focus principally on four research areas:
(1) understanding network structure,
(2) understanding network dynamics,
(3) predictive modeling and simulation for complex networked systems, and
(4) design, situational awareness and control of complex networks.
The program elements consist of a group of Complex Networked Systems Research
Institutes (CNSRI), tightly coupled to an associated individual-investigator-based
Complex Networked Systems Basic Research (CNSBR) program. The CNSRIs will be
principally located at the DOE National Laboratories and are responsible for
identifying research priorities, developing and maintaining a networked systems
modeling and simulation software infrastructure, operating summer schools,
workshops and conferences and coordinating with the CNSBR individual
investigators. The CNSBR individual investigator projects will focus on specific
challenges for networked systems. Relevancy of CNSBR research to DOE needs will
be assured through the strong coupling provided between the CNSBR grants and the
CNSRIs.

COMPLEX NETWORKED SYSTEMS 1

Modeling, Simulation and Analysis of

Complex Networked Systems
Scientists and engineers have long used mathematical and computational models for
the analysis and design of physics-based systems, such as those describing the
evolution of weather and climate, the behavior of complex physical processes in
combustion devices, materials design, weapons systems and energy production.
These models can often be described by systems of partial differential equations,
and sophisticated methodologies have been developed to translate the physics into
mathematical and computational models that can be analyzed and understood to
provide predictive understanding of their behavior. However, there are many
complex systems of importance to the U.S. Department of Energy (DOE) that are
instead described by networks of discrete components. Such systems are not
naturally modeled by systems of partial differential equations, and the required
mathematics is both fundamentally different from what DOE has historically
pursued, and to a large extent is much less well-developed1,2,3.
Examples of networks closely tied to DOEs most critical missions include the
electric power grid and the information networks associated with both large-scale
open science research projects in the Office of Science and with the National
Nuclear Security Administrations Stockpile Stewardship program4. Understanding
and developing a capability to simulate the behavior of the nations power grid at
scale is central to emerging efforts to expand and modernize its operations. The
distributed, collaborative teams that perform open science research must be able to
reliably and securely share significant amounts of information among institutions
and over networks
around the world in
IMPROVED UNDERSTANDING OF COMPLEX NETWORKED
order to enable and
SYSTEMS WILL HELP ANSWER THESE QUESTIONS AND
INFORM POLICY
promote scientific
discovery for the DOE
Can we use understand and quantify the efficacy of
and the nation. Assuring
new security approaches for computer networks?
the safety and reliability
Can we improve the design of computer or
of the nations nuclear
communication networks to be more robust against
partial
failures or intentional attacks?
stockpile is

Can
we
detect a potential catastrophic failure on the
fundamentally an
power grid and prevent or mitigate its effects?
information enterprise
Can we understand how populations will respond to the
that depends on assuring
availability of new energy sources, or to changes in
the underlying complex
energy policy?
Can we understand the behavior of large-scale
information networks.
software systems at a level that allows us to detect
In addition to cyber
subtle vulnerabilities to attack?
networks, such as the
internet and local area
networks, other complex

COMPLEX NETWORKED SYSTEMS

networks whose improved understanding will have significant impact on federal

scientific missions and public policy include social networks, biological networks,
networks describing the spread of disease and even networks or graphs that
describe the organization and execution of very complex scientific software.
The networks of interest to DOE are often very large, containing hundreds of
millions of elements or more, and can range from relatively static in structure (e.g.
the power grid), to rapidly changing and evolving (e.g. the internet, or wireless
networks). Often, as a practical matter, these networks cannot be studied directly.
For example, it is not possible to directly test the effect of controlled shut-down
procedures for segments of the power grid, the spread of a computer virus, or the
spread of a disease vector under different vaccination options without having
potentially drastic economic, safety or health impacts. For infrastructure networks,
replicating all or even part of the physical infrastructure may be prohibitively
expensive. In both these cases, computational modeling and simulation provides a
safe and cost-effective alternative that can help enormously in developing needed
understanding.
Practical questions that can be answered about complex networked systems:
Mathematical analysis and models of complex networks can be used to
computationally explore many practical questions about the performance and
behaviors of real systems:
What are the critical structures and connections in the network? To
understand a sudden change in the behavior of the electrical grid, subtle
linkages between apparently independent components may be very
important.
What are the important dynamics of the network? Understanding the
spreading rate of a human or computer virus requires an understanding of
the dynamics of the network components but also of the connections.
Computational models of complex network models can help to forecast
behaviors. For example, in large-scale computer networks small changes in
protocol timings may lead to unpredicted large-scale changes in behavior.
Ultimately these models will form the foundation for control and design
optimization of complex network systems. How should power transmission
capacity be allocated to optimize real-time grid performance? Where should
network flow sensors be placed to minimize bandwidth requirements for
real-time activity monitoring?
DOE must continue the development of next-generation complex networked
systems that are more secure, less brittle to unexpected events, and more
controllable. For these emerging efforts to be successful, it is essential that a firm
intellectual foundation be provided for understanding and simulating large-scale
networks. Without a strong foundation, DOE will entertain a significant element of
risk associated with the lack of essential knowledge for effectively improving
network performance and understanding potential modes of failure. DOE has a clear
opportunity to exploit its unique set of capabilities in computational and system

COMPLEX NETWORKED SYSTEMS 3

science at massive scale towards this unique and timely opportunity to establish a
strong mathematical and scientific basis to reduce risk associated with these
essential DOE systems and facilities.
Opportunities and Challenges
Many aspects of the DOE mission involve complex networked systems. The
Departments most important information, whether for nuclear weapons
stewardship or for high-energy physics research, flows in large-scale networks. As
plans for new levels of cyberdefense move forward, fundamental understanding of
network performance and effects of changes on performance will be needed. The
nations power grid is a complex network. Understanding of its performance and
ability to predict the effects of both intentional modifications and unforeseen events
is limited5,6. Many of the most important challenges relate to the interactions of
multiple networks. For example, in the modern power grid the flow of information
interacts strongly with the flow of power. In addition, the interaction of evolving
social networks with other network classes is critical to understanding human
generated failure modes. Development of new mathematical and computational
approaches, methods, and tools will play important roles in emerging DOE
programs in these and other areas.
Foundational mathematical and computational research questions:
Addressing these and other issues involving complex networked systems will
require new foundational mathematical and computational research that can
answer the following technical questions:

How can the structure of a complex network be reconstructed

from necessarily limited observations? There are often critical but unknown
interactions among network elements that can create surprising or
emergent behaviors. A fundamental understanding of observability and
reconstructability of a network from limited and imperfect observations is
needed.
How do dynamic processes on the network evolve? Interactions
between network elements on multiple space and time scales can lead to
emergent behaviors. Which interactions are important to understanding
large-scale behavior? What model representations are best for network
dynamics at large scale?
What are the limits to prediction of complex network behavior?
Given imperfect system models and limited computational resources can
well-understood bounds be placed on behaviors? How can these predictive
models be validated in real systems?

For complex, interconnected systems, exhibited behavior that would not be

expected or predicted by inspection of the individual components is often referred
to as emergent.

COMPLEX NETWORKED SYSTEMS

Can simulation methods be scaled to performance levels allowing

their use in design and control optimization for large-scale systems?
Ensembles of simulations to explore classes of behavior as design and
environmental parameters of the model are varied require efficient and
scalable simulation tools.

Research Program Elements

Over the past decade, a significant amount of research has been aimed at
understanding complex networks and their structural characteristics7,8,9,10. By
analogy with the identification of invariant physical laws for physical systems, the
research has focused primarily on the identification of important invariant
properties including degree distribution, clustering coefficient distributions,
shortest-path distribution and connected components. Understanding the behavior
of real complex networks of interest to the DOE requires combining such theoretical
studies with the examination of the actual networks themselves, or with accurate
computational models of these networks whose structural properties can then be
investigated experimentally11. Generating such network models can in itself be an
enormous computational task. This is because of the need to potentially represent
the networks in terms of very large graphs, and because the models may require
generation of extensive synthetic data to provide effective realism. In addition,
these networks must be represented at different levels of fidelity. Some studies
may require that the fundamental elements of the network be described in great
detail, while others may need to exchange fidelity with execution speed. Lower
fidelity models that run quickly can be used to support real-time operational or
control capabilities for a network, or to make very large statistical studies of
network behavior. A full multi-scale approach may also be needed, where some
elements are modeled at very high fidelity while coupled to reduced-order models
that provide a representation of the wider environment.
The ultimate objective of a research program on mathematical modeling, simulation
and analysis of complex networked systems is to develop capabilities for
understanding, designing and control of those networked systems. To do this
requires addressing four major elements: understanding network structure,
understanding network dynamics, predictive modeling and simulation of dynamic
networks, and situational awareness and control. Some of the emerging research
topics in each of these four research areas are described below.
Research area 1: Understanding network structure
To understand network behavior, the underlying topology of the network must be
understood. In general the network systems of interest are of sufficient scale and
complexity that direct measurement of structure will be impossible. Instead
statistical inference and machine learning methods will be used to infer the network
structure indirectly.

COMPLEX NETWORKED SYSTEMS 5

The fundamental problem is
COMPLEX NETWORKED SYSTEMS
to infer the structure of a
RESEARCH PROGRAM ELEMENTS
complex network from a
Understanding network structure
limited set of noisy
observations. The
Understanding network dynamics
observations are often only
indirectly connected to the
Predictive modeling and simulation
basic network structures and
Situational awareness, design and control of
interconnections. In many
complex networks
cases the network may be
spatially distributed (e.g. the
electric grid or Internet) and the scale may be very large. Some of the specific
mathematical and computational challenges include:

Machine learning on general graph structures of particular importance are

learning with relational data, learning with limited or no training data,
hierarchical structure models, and learning with noisy and missing data12;
Current research in complex network structures has mainly focused on
graphs with a single node and link type. To extend the models to more
complex systems graphs with multiple types of nodes and multiple classes of
relations must be considered. In this case many of the basic approaches to
network analysis, such as weighted random walks for similarity metrics,
break down and new methods must be developed;
Development of an understanding of the fundamental limits of network
reconstruction will provide an important foundation for network models.
Under what conditions of sampling and observation errors can network
structures be reconstructed and how should confidence levels be assigned?
Many of these networks are very large. Methods for analyzing the network
structure at varying levels of resolution such as scalable methods for
evaluating eigenvalue spectra of large graphs are needed13;
Methods for analyzing measurements from large networks and evaluating
competing classes of models and inference methods will be required. Work
to scale up statistical data analysis tools on high-performance parallel
computer systems to multi-terabyte or petabyte-class data sets will provide
needed research tools.

Research area 2: Understanding network dynamics

In general the network structure discussed above serves as the substrate for
dynamic processes that are the principal objects of interest. The objective is to
understand behavior under varying conditions at steady state and as the network
evolves, but particularly large-scale emergent behavior changes due to changes or
events in the system or its environment14,15.

COMPLEX NETWORKED SYSTEMS

Graph models are used extensively to model network structure. How can
these mathematical structures extend to incorporate dynamics? How can
they scale as the networks grow?
Modeling dynamic processes in complex networks will require machine
learning methods incorporating multivariate time-varying statistics16.
Understanding the performance of these approaches in noisy, nonstationary
environments raises many open mathematical problems. The learning
process itself becomes a dynamic system coupled to the system being
modeled.
Emergent behavior is one of the hallmarks of complex network dynamics17.
Capturing the nonlinear interactions that produce these effects is an
important component of the network model. Developing stability conditions
and understanding phase transitions will be required.
Understanding the foundations of network dynamics requires providing
theoretical bounds on reconstructability of network dynamics and estimates
of confidence in model structure and parameters.

Research area 3: Mathematical modeling and simulation

Given a mathematical or statistical model of a complex network, simulations can be
developed to realize the model structure and produce dynamic network
behaviors18. Multiple levels of fidelity are needed at the highest resolution levels
the simulations can emulate the bits and instructions of real complex dynamic
networked systems while at lower levels sophisticated mathematical models will be
required. Accurate and validated simulations of complex dynamic networks will be
one of the most important deliverables for DOE programs focused on the
applications of complex networks. Areas of research focus include:

Multi-scale information system simulations that treat parts of the simulated

system at very high fidelity through emulation or virtualization approaches
while simulating the larger parts at lower fidelity. Multiple time scales will
also be important for many types of networks, e.g. the Internet evolves much
more quickly than the power grid;
Highly parallel discrete-event simulations that are the foundation of these
simulation tools;
Performance optimization of large-scale simulations;
Integration of models for human or social components with complex
network simulations;
Analysis of stability, convergence properties and errors in complex network
simulations;
Confidence measures and validation of simulations with ground truth in real
network systems.

COMPLEX NETWORKED SYSTEMS 7

Research area 4: Situational awareness, design and control of complex

networks
The ultimate goal of modeling and simulation research is to develop methods for
understanding and controlling events and anomalies in complex network systems,
which will depend upon progress in research areas 1-3 described above. When online measurements are used to build models and model predictions are, in turn,
used to modify the network operation new dynamics are introduced into this new
closed-loop system. The final research area focuses on the specific research
required for on-line learning in complex networks and the use of control methods to
modify network behavior.

Real-time adaptive machine learning in networks expands the methods

discussed above to work in a real-time on-line environment. The learning
methods must adapt to a nonstationary environment and, in many cases,
scale to very high speed performance.
Distributed situational awareness systems will use multiple sensors
throughout a physically distributed network to build a model of the network
structure and dynamics. Distributing the inputs, the model itself, and the
analysis agents that build it is the focus of this area.
Because of the huge volumes of streaming data often involved in complex
networks, the data must be processed in real time, thus requiring analysis
methods that maintain minimal state information. Mapping sophisticated
learning methods into a streaming environment will be a major challenge.
Many of the problems that these models and simulations will address involve
interaction of the network with some type of adversary. Game-theoretic
models of the adversary integrated into a simulation environment are
required to understand and model these threats19.

COMPLEX NETWORKED SYSTEMS

Program Organizational Structure

The organizational structure of the proposed DOE Office of Science program in
complex networked systems is motivated by the need to rapidly develop the
fundamental science required to make significant advances in our understanding of
these systems. A program that can significantly accelerate the development of
theory, modeling and simulation capabilities can have substantial benefits for the
DOE as it grapples with challenges associated with real complex networked systems
such as the power grid, the internet, and the networks on which the various DOE
enterprises depend. In addition to promoting rapid development of a strong
scientific foundation for understanding complex networked systems, such a
program will be required to develop a substantial base of simulation codes that can
be used to provide scientific understanding and discovery capabilities. This
requirement strongly supports the involvement of the DOE national laboratories,
since collectively they provide substantial experience in the development and
maintenance of large simulation code infrastructures that support various DOE
mission elements.
An effective program will have two main elements. One is the DOE Complex
Networked Systems Research Institute (CNSRI) program, which will include several
large research institutes whose charter is to advance theory, modeling, and software
infrastructure relevant to large-scale DOE networked systems and which will
become long-term focal points for broad national research efforts in complex
networked systems. The second component is a Complex Networked Systems Basic
Research (CNSBR) Program consisting largely of individual investigator research
projects focused on specific challenges for networked systems. To assure the DOErelevance of research projects supported by the CNSBR program, mechanisms will
be provided to strongly couple those projects with the CNSRIs.
The charter of the CNSRI program element is to accelerate the development of the
strong scientific foundation for complex networked systems needed to support DOE
needs. As part of this charter, the CNSRIs will have the responsibility to identify
priority research focus areas in complex networked systems, and will allocate
resources to individual PIs who may subcontract to the CNSRI, or who are resident
for some period of time at the CNSRI (e.g. academic faculty sabbaticals or summer
research opportunities for faculty and students). These activities will be directed by
the CNSRI steering committee, consisting of the principal investigators of the CNSRI,
representatives from the wider research community, and DOE Office of Science
program management.
In addition to their oversight and direction roles, the CNSRIs will be responsible for
the design, development and validation of the significant software infrastructure
supporting full-scale integrated system models of DOE-relevant complex networked
systems. The unique value added by DOE to a research program of this type is the
ability to integrate deep research components into large-scale system applications
something that is well-beyond individual investigator driven research. The critical
importance to DOE of developing these large-scale system applications supports the

COMPLEX NETWORKED SYSTEMS 9

requirement that these Institutes be located at the DOE National Laboratories. Each
CNSRI would focus on an end-to-end application area, for example, modeling and
simulating the dynamic behavior of the national power distribution grid at full scale.
The focus areas for the CNSRIs will demand computational, measurement, and
analytic facilities at a National Lab scale.

Figure 1: The DOE Complex Networked Systems program consists of two main elements, the
Networked Systems Research Institute (CNSRI) program, and the Networked Systems Basic
Research (CNSBR) program. The CNSRI' s are DOE Lab-directed institutes that coordinate
large program sub-areas, while the CNSBRs are individual investigator research efforts based
largely in academia and industry. To assure DOE-relevance, each of the CNSBRs is affiliated
with one or more of the CNSRIs.

The CNSRIs will also run summer research programs, workshops and conferences
that will bring PIs from the CNSBR program, and other federal or international
programs, to promote exchange of ideas and to encourage the development of
collaborative partnerships.
Complementing the integrative function of the DOE Laboratory-based CNSRIs, the
finer-grained CNSBR program is an essential component of the complex network
research program. These projects will perform investigations in focused areas of
complex network research. Each project in the CNSBR program will be affiliated
with one of the CNSRIs whose responsibility it is to integrate the research effort
into the full-scale program.
This program structure also provides natural interfaces to other DOE programs in
the specific focus areas. For example a CNSRI on Information Network Simulation
would provide powerful tools and capabilities to DOE and NNSA cybersecurity and
networking programs.

10

COMPLEX NETWORKED SYSTEMS

The future of complex network theory research at DOE

DOE has a significant opportunity to develop next-generation complex networked
systems that are more secure, less brittle to unexpected events, and more
controllable. The establishment of a new and aggressive basic research program in
modeling, simulation and analysis of complex networked systems will be an
essential enabler of these developments. The resulting strong intellectual
foundation will enable DOE to significantly reduce the risk associated with the lack
of essential knowledge for effectively improving network performance and
understanding potential modes of failure.
With sufficient federal support, a program that couples a group of Complex
Networked Systems Research Institutes with a strong program of individualinvestigator-based basic research projects can be developed over a very few years
that will meet these needs. This program will provide a strong foundation for rapid
progress in the development of understanding and simulation capabilities for largescale complex networks. Research efforts in this program will advance our
understanding of network structure, network dynamics, develop predictive
modeling and simulation capabilities, and provide tools for the design and control
optimization of complex networks. The development of strong foundations in these
four areas will be essential for, e.g., understanding and quantifying the efficacy of
new security approaches for computer networks, improving the design of computer
or communication networks to be more robust against failures or attacks, detecting
potential catastrophic failure on the power grid and preventing or mitigating its
effects, understanding how populations will respond to the availability of new
energy sources or changes in energy policy, and detecting subtle vulnerabilities in
large software systems to intentional attack. Furthermore, it will provide a strong
intellectual foundation that can lead to future breakthroughs in capabilities to build
and operate effective, secure, robust large-scale complex networks in support of the
DOE mission.

COMPLEX NETWORKED SYSTEMS 11

Hendrickson, B. A., and M. H. Wright, Mathematical Research Challenges in

Optimization of Complex Systems, DOE Workshop Report, December 2006.
1

Brown, D. L., et. al., Applied Mathematics at the U. S. Department of Energy: Past,
Present and a View to the Future, Lawrence Livermore National Laboratory Report
LLNL-TR-401536, May 2008.
2

Dunlavy, D. M., Hendrickson, B., and Kolda, T. G., Mathematical Challenges in

Cybersecurity, Sandia Report SAND 2009-0805, February 2009.
3

Catlett, C. , A Scientific Research and Development Approach to Cyber Security,

http://www.science.doe.gov/ascr/ProgramDocuments/Docs/CyberSecurityScience
Dec2008.pdf, December 2008.
4

Carreras, B. A., Lynch, V.E., Dobson I., and Newman, D. E., Critical points and
transitions in an electric power transmission model for cascading failure blackouts,
Chaos 12, 985-994, (2002).

Arionos, S., Bonpard, E., Carbone, A., and Xue, F., Power Grid Vulnerability: A
Complex Network Approach, Chaos 19, 1, March 2009.
6

Amaral, L. A. N., and Ottino, J. M., Complex networks, Eur. Phys. J. B 38, 147-162
(2004).

Newman, M. E. J., The structure and function of complex networks, SIAM Review
45, 167-256 (2003).
8

Albert, R, and Barbabasi, A., Statistical mechanics of complex networks, Reviews

of Modern Physics, 74, 47-97,(2002).

Corogovtsev, A. N., Goltsev, A. V., and Mendes, J.F.F., Critical phenomena in

complex networks, Reviews of Modern Physics, 80, 1275-1335,(2008).
10

Eubank, S., Kumar, V.S.A., Marathe, M. V., and Srinivasan, A., Structural and
algorithmic aspects of massive social networks, in: Discrete Methods in
Epidemiology, Abello J, Cormode, G (ed.). DIMACS Series in Discrete Mathematics
and Theoretical Computer Science, 70, 2006:179-185.
11

Sen, P., Namata, G., Bilgic, M., Getoor, L., Gallagher, B., and Eliassi-Rad, T.,
Collective classification in network data, AI Magazine, 29(3), 93-106 (2008).
12

Aguiar, M. and Bar-Yam, Y., Spectral analysis and the dynamic response of
complex networks, Phys. Rev. E 7, 016106 (2005).

13

12

14

COMPLEX NETWORKED SYSTEMS

Strogatz, S., Exploring complex networks, nature 410, 268, 8 March 2001.

Motter, A., Matias, M., Kurths, J., and Ott, E., Dynamics on Complex Networks and
Applications, Physica D 224, 2006.
15

Sun, J., Tsourakakis, C., Hoke, E., Faloutsos, C., and Eliassi-Rad, T., Two Heads
Better Than One: Pattern Discovery in Time-evolving Multi-aspect Data, Data
Mining and Knowledge Discovery, 17(1), 111-128 (2008).
16

Sole, R. and Valverde, S., Information transfer and phase transitions in a model of
Internet traffic, Physica A 289, 595, 2001.
17

Perumalla, K. and Sundaragopalan, S., High-Fidelity Modeling of Computer

Network Worms, Proc 20th Annual Computer Secuity Applications Conference, 126,
2005.
18

Huberman, B., and Glance, N., Evolutionary games and computer simulations,
Proc. National Academy of Science USA 90, 7716, 1993.
19

LLNL-TR-412733