Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

A Multi-Resolution Dynamic Game Framework for Cross-Echelon Decision-Making in Cyber Warfare

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proposes a two-level dynamic game in which zoom-in and zoom-out operations let a cyber defender refine coarse campaign strategies with tactical detail, lowering attacker value in a five-node case study.

desk verdict The zoom-in/zoom-out framework is cleanly formalized, but the case study's defender improvement is an artifact of overwriting the attacker's strategy rather than a solved equilibrium. read the letter →

arxiv 2507.03021 v1 pith:MS2OBKQF submitted 2025-07-02 cs.CR cs.GT

Ya-Ting Yang , Quanyan Zhu This is my paper
classification cs.CRcs.GT MSC 91A1891A1591A80
keywords cyberwarfaredeceptionmulti-resolutiongameextensive-formMarkovcross-echelondecision-makingpurpleteaming
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Cyber defense operates at two scales at once: the tactical scale of individual attack and deception moves, and the strategic scale of campaign goals and resource allocation. This paper proposes a single dynamic-game model in which each tactical engagement is an extensive-form game tree and the campaign is a zero-sum Markov game whose states are those trees abstracted to coarse nodes. The central claim is that a defender can move between the two scales freely: zooming in replaces a coarse node with its full game tree and re-weights the tree's outcomes by the campaign's continuation values, while zooming out compresses the solved tree back into a coarse policy. In a five-node enterprise-network case study, the paper reports that applying these operations lowers the attacker's expected value compared with a fully abstracted baseline, especially when the attacker is stronger. The paper's contribution is therefore a formal mechanism for choosing modeling resolution selectively, rather than a fixed single-level abstraction.

What carries the argument

The central object is the resolution configuration C ∈ C_S, a per-state choice of whether a base game appears coarsely as s or in detail as Γ^s. The zoom-in operator OP_in reweights the micro game's terminal payoffs using the macro value function (equation (4)) and then solves the micro game for a subgame-perfect equilibrium, i.e., a strategy profile that is a Nash equilibrium in every subgame. The zoom-out operator OP_out compresses the solved micro game back into the coarse world by setting the macro attack probabilities at s equal to the equilibrium outcome probabilities τ^s(z|Φ^s) (equation (5)). Iterating these operators according to a multi-resolution operation plan yields Algorithm 1, which is the concrete mechanism for 'purple teaming'—selectively inspecting one engagement at full detail while keeping the campaign view.

What would settle it

Construct a micro game at some state s with two distinct subgame-perfect equilibria that yield different outcome probabilities, then run the zoom-out rule of equation (5) with each equilibrium and compare the resulting macro policy and value function; if the two values differ, the framework's predictions are not well-defined without an equilibrium-selection rule. A second check is to try adding a terminal outcome 'attacker gains data but stays in state s' to a micro game; under the Section 5.1 assumption such an outcome cannot be represented, so the model should either reject it or require an extended macro graph with internal rewards.

Watch

Extended reading notes

Core claim

The paper claims that cross-echelon cyber decision-making can be captured by nesting two games: micro base games (MBGs), which represent step-by-step attacker-defender interactions as extensive-form game trees, and a macro strategic game (MSG), which treats each MBG as a state in a zero-sum discounted Markov game. Its central contribution is the pair of formally defined zoom-in and zoom-out operators. Zoom-in takes a macro state s represented at coarse resolution and replaces it with the detailed game tree Γ^s, setting each terminal outcome's utility according to the expected continuation value from the macro game under the current macro strategy. Zoom-out takes a micro game that has been solved and sets the macro attack policy at s to the outcome probabilities induced by the micro equilibrium. In the case study, the defender's overall value improves—attacker value falls—when high-resolution tactical insights are folded into the macro plan, with the largest gains against a more capable attacker.

Load-bearing premise

The load-bearing premise is that each micro game's terminal outcomes correspond exactly to the macro graph's outgoing edges, so tactical outcomes that end inside the same macro state (e.g., data exfiltration without advancing) are not representable; the framework also leaves unspecified which subgame-perfect equilibrium is used when zooming out, and different equilibria can give different macro policies.

Editorial extensions

If this is right

  • A defender can solve a coarse campaign game once and then refine only the nodes that matter, avoiding a full fine-grained solve of every subsystem.
  • The case study indicates that the value of zooming depends on attacker capability: against a weak attacker (low λ_A) the coarse baseline may already be enough, while against a capable attacker the longer zoom sequence gives the defender better outcomes.
  • Because the macro game is zero-sum, every reduction in attacker value from a zoom sequence is exactly an improvement in defender payoff, giving a single scalar criterion for evaluating resolution choices.
  • The paper itself notes that zooming is not guaranteed to improve defense; if the coarse policy is already near-optimal, the operations add insight rather than value, so the framework is a way to spend modeling effort selectively rather than a universal improvement guarantee.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves open which subgame-perfect equilibrium is used when a micro game has several; the zoom-out policy in equation (5) can differ across equilibria, so an implementation needs a fixed equilibrium-selection rule to be reproducible.
  • Because the formal setup makes terminal outcomes coincide with outgoing macro edges, any tactical event that ends inside the same macro state—such as data theft that does not advance the attacker—is unrepresentable; an extension would add internal rewards or self-loop outcomes to the macro graph.
  • The choice of which states to zoom could itself be treated as a decision problem: a defender with a budget or time limit could use the macro value function to estimate the value of information of refining each state, a formalization the paper mentions as a future direction.
  • The same zoom-in/zoom-out consistency condition could apply to any hierarchical multi-agent system with coupled local and global dynamics, such as multi-domain operations or interdependent critical infrastructures, which the paper names as generalizing applications.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. This paper proposes a multi-resolution dynamic game framework for cyber warfare decision-making across tactical and strategic echelons. The tactical layer models each node of a network as an extensive-form micro base game (MBG), and the strategic layer is a zero-sum Markov game whose states are abstracted MBGs. The authors define zoom-in and zoom-out operators that update micro payoffs from macro values and update macro attack strategies from micro equilibrium outcome distributions, respectively. A case study on a five-node enterprise network reports that sequential zoom-in/zoom-out operations lower the attacker's game value relative to a completely abstracted game. The abstract claims the framework improves the defender's strategic advantage.

Significance. The paper addresses an important problem—cross-echelon coordination in cyber operations—and provides a clean formal starting point: the MBG and MSG definitions are standard, and the zoom-in operation's use of continuation values is a natural way to couple resolutions. The manuscript also gives a concrete algorithmic description (Algorithm 1) and an illustrative case study, which aids readability. If the effectiveness claim were established, the framework would be a useful tool for interpreting the strategic consequences of tactical detail. However, the core property that the operations improve the defender's strategic outcome is not justified: the zoom-out step writes a non-equilibrium attacker strategy into the macro game, and the evaluation does not verify best-response behavior or a fixed-point condition. These issues are load-bearing for the paper's central claim.

major comments (5)
  1. [§5.2, Eq. (5)] The zoom-out operation rewrites the attacker's macro strategy π_A(·|s) as the micro equilibrium outcome distribution τ^s(·|Φ^s), and the reported values in Figs. 6 and 7 are the expected discounted sums under this imposed profile. In the zero-sum MSG of Definition 8, π_A is an adversarially chosen strategy; no best-response or fixed-point check is performed, so the lower values reflect the payoff of a non-equilibrium profile rather than a strategic improvement for the defender. The manuscript does not provide a theorem bounding the gap between this quantity and the equilibrium value, so the central effectiveness claim of Section 7 is not established.
  2. [§5.1, Eq. (4) and §5.2, Eq. (5)] The micro strategy profile Φ^s used in Eq. (5) is computed from the continuation values V^Π of Eq. (4), which are derived under the pre-zoom macro strategy Π. After Eq. (5) updates π_A, the value function changes; hence Φ^s is generally not an SPNE of the new continuation game. The manuscript does not specify a fixed-point condition or an equilibrium selection rule to break this circularity.
  3. [§5.1, line after Definition 12] The assumption Z^s = {s' | (s,s') ∈ E} identifies every micro terminal outcome with an outgoing macro edge, excluding within-node terminal outcomes such as data exfiltration that ends at the current vertex. These are typical in the motivating cyber scenarios, and the paper does not discuss the modeling consequences of this restriction.
  4. [§6, Figs. 6–7] The case study evaluates Seq3 and Seq5 by comparing values computed under the sequentially updated profile, without re-solving the macro equilibrium after each zoom-out and without comparing against classical abstraction techniques (e.g., state aggregation [28] or action abstraction [21]) on the same topology. The reported improvement is therefore not shown to be robust to the hand-set parameters (λ_A, ν, β) or the single network topology, so the quantitative effectiveness claim is not demonstrated.
  5. [§5.3, Algorithm 1] The paper does not specify which subgame-perfect equilibrium is selected when the micro game has multiple SPNE. Since Eq. (5) uses the outcome distribution τ^s(·|Φ^s), the resulting macro strategy and the reported game values are not well-defined unless an equilibrium selection rule is provided.
minor comments (6)
  1. [§3.2, Eq. (1)] Equation (1) contains a summation over "z ∈ H_z" that is unclear because H_z is not defined; the product already expresses the unique path probability, so please clarify the intended formula.
  2. [§5, paragraph after Definition 12] The text states "2|S| possible resolution configurations," which should presumably be 2^{|S|}; please correct the typesetting.
  3. [§6, Table 1 and accompanying text] The text says the attacker receives a reward of 15 for accessing valuable data at s5, but Table 1 lists ν(s5) = 10; these values should be reconciled.
  4. [§4.1, Definition 8] Definition 8 sets AM_i = E for both players, but the defender's action is described as securing an edge; a short explanation of why the action sets coincide would improve clarity.
  5. [§6, footnote 2] The GAMBiT dataset is cited only by a URL; a full citation with version and access date would improve reproducibility.
  6. [§5.3] The term "purple teaming" is used without definition; a brief explanation or citation would help readers outside the cyber-operations community.

Circularity Check

2 steps flagged · score 6.0 of 10

The reported defender improvement in the case study is generated by the framework's own definitional loop: Eq. (4) builds micro payoffs from the macro value function, Eq. (5) then overwrites the macro attack strategy with the micro outcome distribution, and the resulting value is presented as a strategic advantage without any fixed-point or best-response check.

  1. self definitional [Section 5.1, Eq. (4); Section 5.2, Eq. (5)]
    "given the macro strategy Π, the attacker's utility functions of reaching outcome z ∈ Z^s in the MBG Γ^s are updated as r^s_A(z = s′) = Σ_{a_D∈A^M_D} π_D(a_D|s) Σ_{s′′∈S} [T(s′′|s, a_A=(s,s′),a_D)(R(s,a_A,a_D,s′′)+γV^Π(s′′))] ... the macro attack strategy π_A(·|s) needs to be updated ... π_A(a_A|s)=π(a_A=(s,z)|s)=τ^s(z|Φ^s), ∀z∈Z^s."

    Micro terminal payoffs are defined using the macro value function V^Π of the current macro profile, and the micro equilibrium Φ^s is solved with those payoffs. Then Eq. (5) sets the macro attacker strategy to the resulting micro outcome distribution. Since Section 5.1 defines Z^s = {s′ | (s,s′) ∈ E} and Definition 8 sets A^M_A = E, the objects z and a_A are the same, so Eq. (5) is a definitional identity rather than an independent derivation. The micro equilibrium used for τ^s was computed from the pre-zoom V^Π, but after Eq. (5) updates π_A the value function changes; no fixed-point condition is given. The case study's lower Seq5/Seq3 values are therefore the payoff of this self-consistency loop, not a prediction validated against an independent equilibrium or external benchmark.

  2. fitted input called prediction [Section 7, Discussion and Conclusions; Figs. 6-7]
    "A central contribution of this framework is the introduction of formal zoom-in and zoom-out operations ... Our case study demonstrates that incorporating high-resolution insights into the macro-level planning process can significantly improve the defender's strategic outcomes."

    The effectiveness claim is supported only by comparing the payoff of the profile produced by the zoom operations against the CAG saddle-point value. But the zoom-out step replaces the attacker's macro strategy with the micro outcome distribution, and the value function used in Eq. (4) is the pre-zoom macro value. The reported lower value for Seq5 and Seq3 is an artifact of this substituted attacker profile; the paper does not check that the attacker would play the imposed distribution or that the profile is a saddle-point equilibrium of the updated game. Thus the 'improvement' is fitted by the framework's own construction rather than being a falsifiable strategic prediction.

full rationale

The formal framework itself is largely definitional and internally consistent: Definitions 8-15 define an MSG, zoom-in, and zoom-out without hidden external dependencies. No load-bearing self-citation was found; the cited equilibrium-existence and perfect-recall results (Shapley, Kuhn) are standard external theorems, and the case study uses an external topology dataset only as a graph. However, the headline claim that zoom-in/zoom-out operations 'can significantly improve the defender's strategic outcomes' reduces to the framework's own consistency loop. Eq. (4) constructs micro terminal utilities from the macro value function; Eq. (5) then rebuilds the macro attacker strategy from the resulting micro equilibrium outcome distribution. Because the micro terminal set Z^s is defined to coincide with the macro action space E, this is an equality by construction. The case study compares the value under this rebuilt profile with the CAG equilibrium value and reports the difference as an improvement, but no best-response verification, fixed-point condition, or external benchmark is provided. The improvement is therefore partially circular: the central effectiveness result is produced by the equations that define the multi-resolution operations, not independently established. This warrants a score of 6 rather than a higher score because the framework's definitions and operators, taken as formal constructs, are not themselves circular; only the case-study effectiveness claim is.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The framework's case study rests on hand-chosen scenario parameters (lambda_A, nu, beta, gamma) and a restrictive equivalence between MBG outcomes and macro edges. No free parameters are fitted from data, but the absence of calibration and of reported MBG payoff tables means the central case-study numbers are illustrative. The only genuinely new apparatus, the zoom operators, is defined in terms of the paper's own macro value function and micro equilibria, so it carries an internal consistency burden rather than external validation.

free parameters (5)
  • Attacker capability lambda_A = 0.6 (main case); 0.7 and 0.5 (sensitivity)
    Chosen by hand. It controls the probability that an attacker moves along an unsecured edge in the MSG transition function, so it directly shapes all reported game values.
  • State entry rewards nu(s) = nu(s1)=1, nu(s2)=5, nu(s3)=1, nu(s4)=10, nu(s5)=10
    Chosen by hand in Table 1. These define attacker reward and defender loss per state and are not calibrated to any measurement.
  • Stay penalty beta = -2
    Chosen by hand in Table 1. Penalizes lack of macro progress; the value affects equilibrium policies.
  • Discount factor gamma = unspecified in [0,1]
    Appears in the macro value function and in the zoom-in payoff Eq. (4); no value is given, so the case-study numbers cannot be reproduced from the text.
  • MBG payoff tables for Gamma2-Gamma5 = not reported
    Only the Gamma1 game tree is shown in Figure 5; terminal utilities and tree shapes for the other four micro-base games are omitted, so the case study is under-specified.
assumptions (4)
  • domain assumption Perfect recall (Assumption 1)
    Invoked in Section 3.2 to justify restricting micro strategies to behavior strategies via Kuhn's theorem. Standard but not always realistic for cyber attackers.
  • ad hoc to paper Micro terminal outcomes equal macro edges
    Section 5.1 sets Z^s = {s' | (s,s') in E}, making the zoom operations possible but limiting micro games to choices among macro transitions.
  • domain assumption Macro game is zero-sum with R_D = -R_A
    Section 4.1 models attacker gain as defender loss and defines rewards only from state transitions, excluding non-zero-sum intelligence or deterrence effects.
  • standard math Stationary saddle-point equilibrium exists
    Section 4.2 invokes Shapley's theorem for discounted zero-sum stochastic games, assuming stationarity and consistency between the finite-horizon formulation and the discounted value equations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Multi-Resolution Dynamic Game Framework for Cross-Echelon Decision-Making in Cyber Warfare." pith.science (2026). https://pith.science/paper/MS2OBKQF

@misc{pith2026250703021,
  author       = {Pith},
  title        = {Pith review of: A Multi-Resolution Dynamic Game Framework for Cross-Echelon Decision-Making in Cyber Warfare},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MS2OBKQF}},
  note         = {Machine review of arXiv:2507.03021}
}
read the original abstract

Cyber warfare has become a critical dimension of modern conflict, driven by society's increasing dependence on interconnected digital and physical infrastructure. Effective cyber defense often requires decision-making at different echelons, where the tactical layer focuses on detailed actions such as techniques, tactics, and procedures, while the strategic layer addresses long-term objectives and coordinated planning. Modeling these interactions at different echelons remains challenging due to the dynamic, large-scale, and interdependent nature of cyber environments. To address this, we propose a multi-resolution dynamic game framework in which the tactical layer captures fine-grained interactions using high-resolution extensive-form game trees, while the strategic layer is modeled as a Markov game defined over lower-resolution states abstracted from those game trees. This framework supports scalable reasoning and planning across different levels of abstraction through zoom-in and zoom-out operations that adjust the granularity of the modeling based on operational needs. A case study demonstrates how the framework works and its effectiveness in improving the defender's strategic advantage.

Figures

Figures reproduced from arXiv: 2507.03021 by the authors.

Figure 1
Figure 1. An illustration of multi-layered decision-making (s for lower and Γ for higher resolution). (a) The MSG abstracts the system considered during cyber warfare as a network of interconnected MBGs. (b) The strategic layer governs inter-MBG planning at a coarser resolution, while the tactical layer focuses on detailed action sequences within each MBG. Given the multi-layered nature of decision-making in cyber operations,… view at source ↗
Figure 2
Figure 2. An illustrative example for multi-resolution operations. For instance, consider a set of MBGs of size |S| = 3, and let the initial resolution configuration be C0 = Cl = {s 1 , s2 , s3}, where each state s i ∈ S is initially represented at the macro level (low-resolution), forming a completely abstracted configuration (CAG). Then, given a sequence of resolution operations [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗
Figure 3
Figure 3. An enterprise network for case study. A possible attack path contains five nodes, including the web server in the DMZ site, devices in the IT site, user devices in the human resource or finance site, the devices in the developer site, and the critical asset located on the developer server. such a plan offers is the ability to zoom in on specific parts of the system to exam￾ine whether local interactions, when modele… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The MSG representation constructed from the enterprise network topology in [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Extensive-form game tree for the MBG Γ1 associated with vertex s1, represent￾ing the web server in the DMZ site. Red and blue nodes indicate attacker and defender decision points, respectively, while yellow nodes denote terminal outcomes z ∈ Z1. This structure captures…
Figure 6
Figure 6. Figure 6: The resulting game (vertex) value for each state s i ∈ S of the multi-resolution game from the MSG in [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: The resulting game (vertex) value for each state s i ∈ S of the multi-resolution game is shown for different attacker’s capabilities λA: (a) corresponds to λA = 0.7, and (b) corresponds to λA = 0.5. gies. This is evident in the comparison between Seq5 and Seq3, where S…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Guarding Against Malicious Biased Threats (GAMBiT) Experiments: Revealing Cognitive Bias in Human-Subjects Red-Team Cyber Range Operations

    cs.CR 2025-08 conditional novelty 6.0 of 10

    Three multi-modal datasets capture 59 skilled attackers' full operational traces (keystrokes, shell history, PCAP, surveys) in a simulated enterprise network, with labels designed to reveal cognitive biases.

Reference graph

Works this paper leans on

35 extracted references · 30 canonical work pages · cited by 1 Pith paper

  1. [28]

    In: Pro- ceedings of the 41st IEEE Conference on Decision and Control, 2002

    Ren, Z., Krogh, B.H.: State aggregation in markov decision processes. In: Pro- ceedings of the 41st IEEE Conference on Decision and Control, 2002. vol. 4, pp. 3819–3824. IEEE (2002)

    work page 2002

  2. [21]

    In: Proceedings of the AAAI Con- ference on Artificial Intelligence

    Marino, J.R., Moraes, R.O., Toledo, C., Lelis, L.H.: Evolving action abstractions for real-time planning in extensive-form games. In: Proceedings of the AAAI Con- ference on Artificial Intelligence. vol. 33, pp. 2330–2337 (2019)

    work page 2019

  3. [1]

    Daedalus149(2), 133–149 (2020)

    Acton, J.M.: Cyber warfare & inadvertent escalation. Daedalus149(2), 133–149 (2020)

    work page 2020

  4. [2]

    IEEE Communications Surveys & Tutorials21(2), 1851–1877 (2019)

    Alshamrani, A., Myneni, S., Chowdhary, A., Huang, D.: A survey on advanced persistent threats: Techniques, solutions, challenges, and research opportunities. IEEE Communications Surveys & Tutorials21(2), 1851–1877 (2019)

    work page 2019

  5. [3]

    SIAM (1998)

    Başar, T., Olsder, G.J.: Dynamic noncooperative game theory. SIAM (1998)

    1998

  6. [4]

    Electricity information sharing and analysis center (E-ISAC)388(1-29), 3 (2016)

    Case, D.U.: Analysis of the cyber attack on the ukrainian power grid. Electricity information sharing and analysis center (E-ISAC)388(1-29), 3 (2016)

    work page 2016

  7. [5]

    Foreign Aff.75, 37 (1996)

    Cohen, E.A.: A revolution in warfare. Foreign Aff.75, 37 (1996)

    work page 1996

  8. [6]

    Quantitative Economics3(1), 53–93 (2012)

    Doraszelski, U., Judd, K.L.: Avoiding the curse of dimensionality in dynamic stochastic games. Quantitative Economics3(1), 53–93 (2012)

    work page 2012

Show all 35 references
  1. [7]

    Survival53(1), 23–40 (2011)

    Farwell, J.P., Rohozinski, R.: Stuxnet and the future of cyber war. Survival53(1), 23–40 (2011)

    2011

  2. [8]

    Springer Science & Business Media (2012)

    Filar, J., Vrieze, K.: Competitive Markov decision processes. Springer Science & Business Media (2012)

    2012

  3. [9]

    In: International Conference on Decision and Game Theory for Security

    Ge, Y., Zhu, Q.: Mega-pt: A meta-game framework for agile penetration testing. In: International Conference on Decision and Game Theory for Security. pp. 24–44. Springer (2024)

    2024

  4. [10]

    In: 2020 12th international conference on cyber conflict (CyCon)

    Hartmann, K., Giles, K.: The next generation of cyber-enabled information war- fare. In: 2020 12th international conference on cyber conflict (CyCon). vol. 1300, pp. 233–250. IEEE (2020)

    2020

  5. [11]

    In: Interna- tionalConferenceonDecisionandGameTheoryforSecurity.pp.273–294.Springer (2017)

    Horák, K., Zhu, Q., Bošansk` y, B.: Manipulating adversary’s belief: A dynamic game approach to deception by design for proactive network security. In: Interna- tionalConferenceonDecisionandGameTheoryforSecurity.pp.273–294.Springer (2017)

    2017

  6. [12]

    Comput- ers & Security p

    Javadpour, A., Ja’fari, F., Taleb, T., Shojafar, M., Benzaïd, C.: A comprehensive survey on cyber deception techniques to improve honeypot performance. Comput- ers & Security p. 103792 (2024)

    2024

  7. [13]

    Defense & Security Analysis35(2), 147–169 (2019)

    Johnson, J.: Artificial intelligence & future warfare: implications for international security. Defense & Security Analysis35(2), 147–169 (2019)

    2019

  8. [14]

    John Wiley & Sons (2021)

    Kamhoua, C.A., Kiekintveld, C.D., Fang, F., Zhu, Q.: Game theory and machine learning for cyber security. John Wiley & Sons (2021)

    2021

  9. [15]

    EPJ Quantum Tech- nology 8(1), 24 (2021)

    Krelina, M.: Quantum technology for military applications. EPJ Quantum Tech- nology 8(1), 24 (2021)

    2021

  10. [16]

    Econometrica 50(4), 863–894 (1982), http://www.jstor.org/stable/1912767

    Kreps, D.M., Wilson, R.: Sequential equilibria. Econometrica 50(4), 863–894 (1982), http://www.jstor.org/stable/1912767

    1982

  11. [17]

    Contributions to the Theory of Games2(28), 193–216 (1953) 20 Ya-Ting Yang and Quanyan Zhu

    Kuhn, H.W.: Extensive games and the problem of information. Contributions to the Theory of Games2(28), 193–216 (1953) 20 Ya-Ting Yang and Quanyan Zhu

    1953

  12. [18]

    arXiv preprint arXiv:1301.0580 (2012)

    Lagoudakis,M.,Parr,R.:Valuefunctionapproximationinzero-summarkovgames. arXiv preprint arXiv:1301.0580 (2012)

    2012 arXiv

  13. [19]

    arXiv preprint arXiv:2403.10570 (2024)

    Li, T., Zhu, Q.: Symbiotic game and foundation models for cyber deception oper- ations in strategic cyber warfare. arXiv preprint arXiv:2403.10570 (2024)

    2024 arXiv

  14. [20]

    World Scientific News 190(1), 1–69 (2024)

    Mallick, M.A.I., Nath, R.: Navigating the cyber security landscape: A compre- hensive review of cyber-attacks, emerging trends, and recent developments. World Scientific News 190(1), 1–69 (2024)

    2024

  15. [22]

    Cambridge University Press (2020)

    Maschler, M., Zamir, S., Solan, E.: Game theory. Cambridge University Press (2020)

    2020

  16. [23]

    Optimiza- tion Methods and Software14(1-2), 99–116 (2000)

    Nash, S.G.: A multigrid approach to discretized optimization problems. Optimiza- tion Methods and Software14(1-2), 99–116 (2000)

    2000

  17. [24]

    In: Proceedings of the 7th European Conference on Informa- tion Warfare

    Ottis, R.: Analysis of the 2007 cyber attacks against estonia from the information warfare perspective. In: Proceedings of the 7th European Conference on Informa- tion Warfare. p. 163. Academic Publishing Limited Reading, MA (2008)

    2008

  18. [25]

    Econometrica69(5), 1261–1281 (2001)

    Pakes, A., McGuire, P.: Stochastic algorithms, symmetric markov perfect equilib- rium, and the ‘curse’of dimensionality. Econometrica69(5), 1261–1281 (2001)

    2001

  19. [26]

    Powell, W.B.: Approximate Dynamic Programming: Solving the curses of dimen- sionality, vol. 703. John Wiley & Sons (2007)

    2007

  20. [27]

    Journal of Cyber Security Technology9(1), 28–78 (2025)

    Radanliev, P.: Cyber diplomacy: defining the opportunities for cybersecurity and risks from artificial intelligence, iot, blockchains, and quantum computing. Journal of Cyber Security Technology9(1), 28–78 (2025)

    2025

  21. [29]

    Com- puters & security49, 70–94 (2015)

    Robinson, M., Jones, K., Janicke, H.: Cyber warfare: Issues and challenges. Com- puters & security49, 70–94 (2015)

    2015

  22. [30]

    Proceedings of the national academy of sciences 39(10), 1095–1100 (1953)

    Shapley, L.S.: Stochastic games. Proceedings of the national academy of sciences 39(10), 1095–1100 (1953)

    1953

  23. [31]

    In: Technical report

    Strom, B.E., Applebaum, A., Miller, D.P., Nickels, K.C., Pennington, A.G., Thomas, C.B.: Mitre att&ck: Design and philosophy. In: Technical report. The MITRE Corporation (2018)

    2018

  24. [32]

    In: 2016 8th International Conference on Cyber Conflict (CyCon)

    Tabansky, L.: Towards a theory of cyber power: The israeli experience with in- novation and strategy. In: 2016 8th International Conference on Cyber Conflict (CyCon). pp. 51–63 (2016). https://doi.org/10.1109/CYCON.2016.7529426

    2016

  25. [33]

    Journal of Systems Science and Complexity30(1), 46–67 (2017)

    Wen, G., Yu, W., Yu, X., Lü, J.: Complex cyber-physical networks: From cyberse- curity to security control. Journal of Systems Science and Complexity30(1), 46–67 (2017)

    2017

  26. [34]

    arXiv preprint arXiv:2505.21244 (2025)

    Yang, Y.T., Zhu, Q.: When to deceive: A cross-layer stackelberg game framework for strategic timing of cyber deception. arXiv preprint arXiv:2505.21244 (2025)

    2025 arXiv

  27. [35]

    In: Decision and Game Theory for Security: 9th International Conference, GameSec 2018, Seattle, WA, USA, October 29–31, 2018, Proceedings 9

    Zhang, T., Zhu, Q.: Hypothesis testing game for cyber deception. In: Decision and Game Theory for Security: 9th International Conference, GameSec 2018, Seattle, WA, USA, October 29–31, 2018, Proceedings 9. pp. 540–555. Springer (2018)

    2018

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.