Abstraction Methods for Solving Graph-Based Security Games

Basak, Anjon; Fang, Fei; Nguyen, Thanh H.; Kiekintveld, Christopher

doi:10.1007/978-3-319-46840-2_2

“…(Kiekintveld et al 2009;Schlenker et al 2016); or game abstraction -e.g. (Wang et al 2018;Basak et al 2016). Utilization of these techniques requires tailoring a solution method to characteristic game proper-ties, what leads to a highly efficient, though game-dedicated algorithm.…”

Section: Related Workmentioning

confidence: 99%

Double-Oracle Sampling Method for Stackelberg Equilibrium Approximation in General-Sum Extensive-Form Games

Karwowski

¹

,

Mańdziuk

²

2020

AAAI

View full text Add to dashboard Cite

The paper presents a new method for approximating Strong Stackelberg Equilibrium in general-sum sequential games with imperfect information and perfect recall. The proposed approach is generic as it does not rely on any specific properties of a particular game model. The method is based on iterative interleaving of the two following phases: (1) guided Monte Carlo Tree Search sampling of the Follower's strategy space and (2) building the Leader's behavior strategy tree for which the sampled Follower's strategy is an optimal response. The above solution scheme is evaluated with respect to expected Leader's utility and time requirements on three sets of interception games with variable characteristics, played on graphs. A comparison with three state-of-the-art MILP/LP-based methods shows that in vast majority of test cases proposed simulation-based approach leads to optimal Leader's strategies, while excelling the competitive methods in terms of better time scalability and lower memory requirements.

show abstract

“…Kiekintveld et al (2009); Schlenker et al (2016); or game abstraction -e.g. Wang et al (2018); Basak et al (2016). Utilization of these techniques requires tailoring a solution method to characteristic game properties, what leads to a highly efficient, though game-dedicated algorithm.…”

Section: Related Workmentioning

confidence: 99%

Double-oracle sampling method for Stackelberg Equilibrium approximation in general-sum extensive-form games

Karwowski,

Mańdziuk

2019

Preprint

0

View full text Add to dashboard Cite

The paper presents a new method for approximating Strong Stackelberg Equilibrium in general-sum sequential games with imperfect information and perfect recall. The proposed approach is generic as it does not rely on any specific properties of a particular game model. The method is based on iterative interleaving of the two following phases: (1) guided Monte Carlo Tree Search sampling of the Follower's strategy space and (2) building the Leader's behavior strategy tree for which the sampled Follower's strategy is an optimal response. The above solution scheme is evaluated with respect to expected Leader's utility and time requirements on three sets of interception games with variable characteristics, played on graphs. A comparison with three state-of-the-art MILP/LP-based methods shows that in vast majority of test cases proposed simulation-based approach leads to optimal Leader's strategies, while excelling the competitive methods in terms of better time scalability and lower memory requirements.

show abstract

“…Instead of using the double oracle algorithm, there are several works which focus on using abstraction techniques to reduce the strategy space size in structured security game [67][68][69]. Typically, abstraction techniques exploit action space similarity to build a simpler problem and solve it.…”

Section: Stackelberg Security Game On Structured Domainsmentioning

confidence: 99%

“…The abstraction can be lossy or lossless depending on the problem. For example, Basak et al build a small graph with only key nodes on the original graph to compute the defender's patrol strategy on the small graph and then mapping the solution, i.e., paths from key nodes, to the original by selecting the shortest path between key nodes on the original graph [68].…”

Section: Stackelberg Security Game On Structured Domainsmentioning

confidence: 99%

“…Abstraction has been successfully applied to solve large-scale games by exploiting the intrinsic similarity of strategy space to shrink the game to a simpler one before solving it. As both players take paths as their strategies rather than a specific node and the zones on the attacker's escape path play a vital role for the defender to detect the attacker, the methods proposed in [68,69,99], which typically remove the nodes unnecessary to reduce the size of the game, cannot be directly applied to our model. The basic idea of the abstraction method applied here is that the defender can only determine the coverage of a large area, which can be assigned to small areas following a pre-defined mapping method, which ensures that the generated coverage satisfies the flow conservation constraints.…”

Section: Abstraction To Further Improve Scalabilitymentioning

confidence: 99%

See 1 more Smart Citation

Defending on networks : applying game theory to prevent illegal activities in structured security domains

Wang¹

0

View full text Add to dashboard Cite

Pursuing a PhD degree is definitely the most important decision I have made in my past life. The four-year study teaches me how to be a professional researcher to tackle the problems, a grown-up to work with other people and a mature guy to get along with myself. I have learnt from many brilliant people, who taught, helped and inspired me.It is time to express my sincere gratitude to them.First of all, I want to thank my supervisor Bo An. Without your supervision, this point can not be reached. Thank you for giving me the opportunity to pursue my PhD degree under your supervision. Four years ago, I joined your group without much knowledge about research. Thanks for your endless patience, insightful guidance and constant encouragement during every step of doing research. I am also deep impressed and motivated by your enthusiasm and profession to the research and those will be the fundamental things I will strive for in my future career. You reshapes my understanding about the research with your commitment to work on problems with impacts to the real-world and benefits to the society. Your critical comments always deepen my understanding on the research problems and the discussions with you always benefit me a lot.Not only on about research, I also learn a lot about the principles for the life: the way to dealing with failures because "rejection is a part of your academic life", the willingness to chase perfection and excellence because we always fight until the last minute of the conference deadlines and the love to Sichuan cuisine. Let me quote that "you set the pace and I will try to adjust to you". I will be forever grateful to your supervision and I hope I can let you be proud of me someday in the future.I would like to thank my Thesis Advisory Committee (TAC) members: Gao Cong, Rongxing Lu, Yilin Mo and Xiaohui Bei. Those discussions help me to improve the vii viii works and build a better vision of my research. Your guidance are valuable for me and I will always be thankful for your efforts in advising me.

show abstract

Abstraction Methods for Solving Graph-Based Security Games

Cited by 8 publications

References 21 publications

Double-Oracle Sampling Method for Stackelberg Equilibrium Approximation in General-Sum Extensive-Form Games

Double-Oracle Sampling Method for Stackelberg Equilibrium Approximation in General-Sum Extensive-Form Games

Double-oracle sampling method for Stackelberg Equilibrium approximation in general-sum extensive-form games

Defending on networks : applying game theory to prevent illegal activities in structured security domains

Contact Info

Product

Resources

About