Carnegie Mellon University
Browse

Attack and Defense Strategies in Cyber-Physical Systems with Varying Levels of System and Opponent Knowledge

Download (3.64 MB)
thesis
posted on 2015-05-01, 00:00 authored by Bruce Debruhl
Advances in computing, communications, and sensing have enabled exciting opportunities for
large scale applications of cyber-physical systems (CPS) to energy, transportation, healthcare,
and defense. All of these services support critical applications, making CPS security crucial.
For example, an attack against the smart-grid, or a power grid enhanced with CPS, may result in
devastating regional blackouts. Fortunately, the technologies that enable CPS allows us to design
attack and defense strategies leveraging robust sensing and actuation.
In this thesis, we explore the interaction of two adversarial players with a shared cyberphysical
system. We investigate how a player with limited information about the CPS or their
opponent chooses an attack or defense. In particular, we explore the following question: how is
an agent’s strategy affected by the amount of knowledge they have about the CPS they interact
with and their opponent’s strategy?
We consider various scenarios to explore this problem including: an agent that interacts
with a known system and known opponent, an agent that interacts with a known system and an
opponent with assumed behavior, an agent that interacts with a known system and an unknown
opponent, and an agent that interacts with a known opponent and a partially known system. For
each of these scenarios we provide a proof-of-concept attack or defense to demonstrate security
challenges and opportunities. We also introduce other scenarios based on system and opponent
knowledge levels that demonstrates exciting future research opportunities.

History

Date

2015-05-01

Degree Type

  • Dissertation

Department

  • Electrical and Computer Engineering

Degree Name

  • Doctor of Philosophy (PhD)

Advisor(s)

Patrick Tague

Usage metrics

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC