a game theoretic study of attack and defense in cyber-physical systems
DESCRIPTION
A Game Theoretic Study of Attack and Defense in Cyber-Physical Systems. Chris Y. T. Ma, Nageswara S. V. Rao, David K. Y. Yau. Agenda. Motivation System model Boolean attack and defense System with robustness Conclusion. Motivation. Cyber-physical systems - PowerPoint PPT PresentationTRANSCRIPT
A Game Theoretic Study of Attack and Defense in Cyber-Physical Systems
Chris Y. T. Ma, Nageswara S. V. Rao, David K. Y. Yau
Agenda
Motivation System model Boolean attack and defense System with robustness Conclusion
Motivation
Cyber-physical systems Model a number of engineering infrastructure systems
Physical – hardware components Cyber – computations, communications
Susceptible to attacks
Motivation
Our objectives Use of game theoretic formulations to capture the
attack and defense of cyber-physical systems Study the survival of the cyber-physical systems using
different utility functions
Motivation
Our observations Pure strategy Nash Equilibrium (NE) may not exist Cost boundaries (budgets) may determine the NE
outcome The presence of NE does not mean the system
survives
System Model
System Model
Boolean Attack and Defense
Special case of attacks where the cyber and physical parts can be attacked or defended as whole units
Successful attack on either cyber or physical part will disrupt the whole system
Boolean Attack and Defense
Boolean Attack and Defense
System with Robustness
General case when resources are not represented as one whole unit
Consider different benefit functions
System with Robustness
The players’ best response functions
System with Robustness
General Benefit and Cost Functions
General Benefit and Cost FunctionsOne-space cases
Observation Pure strategy Nash Equilibrium is rare, most likely to
exist when the attacker has tight budget
General Benefit and Cost FunctionsOne-space cases
General Benefit and Cost FunctionsTwo-space cases
Observations Resource allocation is non-trivial even without an
attacker, and greedy approach may be sub-optimal, e.g., the S-shaped benefit function
The NE results are sensitive to the parameters of the payoff functions in the two spaces
General Benefit and Cost FunctionsTwo-space cases
Observations
System ACyber space: Ba
Physical space: Ba
NE: X
System BCyber space: Bb
Physical space: Bb
NE: X
System CCyber space: Ba
Physical space: Bb
NE: ?
General Benefit and Cost FunctionsTwo-space cases
Conclusions
Presented a game theoretic formulation of the interplay between a rational attacker and a rational defender in cyber-physical system security
Studied the presence (or absence) of pure strategy Nash Equilibrium using different payoff functions