arxiv: 2605.09792 · v1 · submitted 2026-05-10 · 💻 cs.CR

Recognition: 2 theorem links

· Lean Theorem

Operationalizing Cybersecurity Governance for Mitigation Planning with Attack-Path Modeling and Reinforcement Learning

Philip Huff , Dakota Dale , Harshith Guduru , Rohan Singh , Qinghua Li

Authors on Pith no claims yet

Pith reviewed 2026-05-12 02:32 UTC · model grok-4.3

classification 💻 cs.CR

keywords cybersecurity governanceMITRE ATT&CKreinforcement learningattack path modelingNIST CSFmitigation planningvariable-order Markov modeladversary simulation

0 comments

The pith

Mapping maturity assessments to adversary techniques lets reinforcement learning select budget-constrained mitigations.

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

This paper shows how to convert abstract cybersecurity governance frameworks into specific defensive actions that respond to realistic adversary behavior under limited resources. It translates NIST CSF maturity scores into MITRE ATT&CK mitigation capabilities and trains a variable-order Markov model on observed technique sequences to simulate likely attack paths. Deep reinforcement learning then optimizes which mitigations to apply, accounting for concurrent threats and explicit costs. The resulting policies prove stable across reward settings and produce interpretable plans tied to an organization's assessed posture. A reader cares because governance standards currently give little help choosing which controls to fund first when threats evolve.

Core claim

The paper claims that linking CSF maturity assessments to ATT&CK mitigation capabilities, then embedding a variable-order Markov model of attack sequences inside a deep reinforcement learning environment, enables the generation of stable, resource-aware mitigation policies that optimize risk reduction while handling multiple concurrent adversaries and realistic costs.

What carries the argument

The central mechanism is the DRL environment that uses the CSF-to-ATT&CK mapping to define available actions and a variable-order Markov model trained on ATT&CK sequences to reconstruct attack paths via beam search for reward calculation and joint optimization under budget constraints.

If this is right

Mitigation plans can be derived directly from standard CSF maturity assessments without manual translation.
Defense choices adapt to observed adversary sequences rather than fixed rules or checklists.
Explicit budget constraints produce measurable cost-risk trade-offs in the output policies.
Stable policies emerge across multiple reward formulations and environment configurations.
Concurrent adversary simulation yields more robust mitigation recommendations than single-path models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same mapping-plus-simulation structure could apply to other governance frameworks by building equivalent technique linkages.
Generated plans might serve as initial inputs for automated security orchestration systems that enforce controls dynamically.
Periodic retraining of the Markov model on fresh incident data could keep recommendations aligned with evolving threats.
Comparison of model outputs against historical breach data would offer an independent check on the adversary simulation accuracy.

Load-bearing premise

The variable-order Markov model trained on ATT&CK sequences accurately captures real adversary behavior, and the CSF-to-ATT&CK mapping faithfully converts organizational maturity into usable mitigation options.

What would settle it

A controlled red-team test that applies the generated mitigation plans yet still records high attack success rates using ATT&CK techniques outside the training sequences would falsify the effectiveness of the resulting strategies.

Figures

Figures reproduced from arXiv: 2605.09792 by Dakota Dale, Harshith Guduru, Philip Huff, Qinghua Li, Rohan Singh.

**Figure 2.** Figure 2: Ablation study results showing learning dynamics across configuration settings, measured by the mean [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗

**Figure 3.** Figure 3: Attack-path reconstruction for School 1 with multiple paths output from the beam search, and the correspond [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗

**Figure 4.** Figure 4: Base percent-of-budget lookup map (PctCost) for ordinal cost and complexity ratings. Color intensity indicates the fraction of the episode budget consumed by a mitigation prior to maturity scaling. D Base Budget Lookup Map [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗

**Figure 5.** Figure 5: Resource-spread map Spread(type, resource) used to estimate per-target resource availability for adversaries. Values represent expected operator-equivalents allocated per target per planning period. We compute Spread from analyst-provided ranges for (i) approximate adversary staffing and (ii) typical monthly targeting volume for that adversary type. The returned value is the ratio of these quantities, expr… view at source ↗

read the original abstract

We address a fundamental challenge in cybersecurity operations of translating governance frameworks into actionable mitigation decisions under realistic resource constraints. Frameworks such as the NIST Cybersecurity Framework (CSF) provide widely adopted measures of organizational maturity, but do not directly support the selection and prioritization of defensive strategies against adversarial behavior. We present a system that operationalizes governance frameworks by mapping CSF maturity assessments into MITRE ATT\&CK mitigation capabilities, which enables direct integration of organizational security posture with adversary-informed defensive planning. To manage adversary complexity, we employ a Variable-Order Markov Model (VOMM) trained on observed ATT\&CK technique sequences to enable scalable adversary simulation within a Deep Reinforcement Learning (DRL) environment. We reconstruct likely attack paths and defensive responses using beam search, and then jointly optimize mitigation selection under explicit budget constraints. Our environment supports concurrent adversaries and realistic mitigation costs. Across multiple reward formulations and configurations, we show that the approach produces stable policies, meaningful cost-risk trade-offs, and interpretable mitigation plans aligned with organizational maturity. These results demonstrate that adversary-aware DRL can generate practical, resource-constrained defense strategies grounded in real-world frameworks and threat behavior.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper sketches an end-to-end pipeline from CSF maturity to ATT&CK mitigations via VOMM simulation and DRL optimization, but supplies no numbers or checks on whether the models match reality.

read the letter

The main takeaway is that the authors have wired together CSF assessments, ATT&CK techniques, a variable-order Markov model for attack sequences, and deep reinforcement learning to pick mitigations under budget limits. The abstract describes mapping maturity levels to available defenses, running beam search on likely paths, and optimizing across concurrent adversaries with realistic costs. That combination is the concrete step they take beyond just listing controls or running generic risk matrices. It does a reasonable job of making the governance framework drive actual decisions while keeping the adversary side scalable through the Markov model instead of full graphs. Handling multiple simultaneous threats and explicit budget constraints also feels like a practical engineering choice rather than an oversimplification. The claims about stable policies and interpretable plans aligned with maturity follow logically from the setup they describe. The soft spots sit at the inputs and the outputs. The abstract asserts meaningful cost-risk trade-offs and stable policies but gives no metrics, baselines, error bars, or even sample results, so there is no way to judge whether the DRL actually beats simpler selection methods. More importantly, nothing in the text checks whether the VOMM trained on ATT&CK sequences produces paths whose statistics match real incidents, or whether the CSF-to-mitigation translation assigns realistic costs and coverage. The stress-test note is accurate on this point: if either piece drifts from reality, the optimized strategies become artifacts of the simulation. The thinking is clear and the components are drawn from established sources, but the work stays at the level of a described system rather than a validated one. This would interest readers working on applied RL for security operations or on turning frameworks into prioritized plans. A practitioner could pull the architecture and try to implement the simulation, but only after adding their own validation. I would send it for peer review so referees can require sensitivity checks on the VOMM fidelity and the mapping accuracy, plus some quantitative comparison to baselines. That would clarify whether the pipeline delivers usable improvements.

Referee Report

2 major / 2 minor

Summary. The paper claims to operationalize the NIST Cybersecurity Framework (CSF) by mapping maturity assessments to MITRE ATT&CK mitigation capabilities, training a Variable-Order Markov Model (VOMM) on observed ATT&CK technique sequences for adversary simulation, using beam search to reconstruct attack paths, and applying Deep Reinforcement Learning (DRL) to jointly optimize mitigation selection under explicit budget constraints and concurrent adversaries. It asserts that the resulting system yields stable policies, meaningful cost-risk trade-offs, and interpretable plans aligned with organizational maturity across multiple reward formulations.

Significance. If the central claims hold with proper validation, the work could meaningfully advance practical cybersecurity governance by providing a concrete, adversary-aware mechanism to translate high-level CSF assessments into prioritized, resource-constrained mitigations. The integration of established frameworks (CSF, ATT&CK) with scalable simulation and RL optimization is a strength that could improve actionability of governance frameworks in operational settings.

major comments (2)

[Abstract] Abstract: the assertion that 'across multiple reward formulations and configurations, we show that the approach produces stable policies, meaningful cost-risk trade-offs' is presented without any quantitative results, tables, figures, convergence metrics, error bars, baseline comparisons, or validation details in the available manuscript text. This directly undermines evaluation of the central claim that adversary-aware DRL generates practical, governance-grounded strategies.
[Model description] Model description (VOMM and CSF-to-ATT&CK mapping): the pipeline assumes the VOMM trained on ATT&CK sequences faithfully reproduces real-world adversary behavior for beam-search path reconstruction and DRL optimization, and that the CSF maturity mapping produces realistic available mitigations with accurate costs. No validation against incident data, cross-checks, sensitivity analysis, or fidelity metrics are reported; deviations here would render the reported policies simulation artifacts rather than evidence of practical utility.

minor comments (2)

Clarify the exact mechanism and any assumptions in the CSF maturity to mitigation availability/cost mapping, including how maturity levels constrain the action space in the DRL environment.
Ensure consistent definition of all acronyms (e.g., VOMM, DRL, CSF) on first use and provide a brief overview of the beam-search parameters used for attack-path reconstruction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. The comments identify opportunities to strengthen the presentation of results and the discussion of modeling assumptions. We respond to each major comment below and indicate the revisions we will incorporate.

read point-by-point responses

Referee: [Abstract] Abstract: the assertion that 'across multiple reward formulations and configurations, we show that the approach produces stable policies, meaningful cost-risk trade-offs' is presented without any quantitative results, tables, figures, convergence metrics, error bars, baseline comparisons, or validation details in the available manuscript text. This directly undermines evaluation of the central claim that adversary-aware DRL generates practical, governance-grounded strategies.

Authors: We agree that the abstract would be strengthened by including specific quantitative indicators. The full manuscript reports experimental outcomes in Sections 5 and 6, including convergence behavior under different reward formulations, policy stability metrics, cost-risk trade-off tables, and comparisons across budget constraints and adversary counts. We will revise the abstract to concisely summarize representative quantitative findings (e.g., stability scores and risk-reduction ranges) so that the high-level claims are directly supported by the empirical content already present in the paper. revision: yes
Referee: [Model description] Model description (VOMM and CSF-to-ATT&CK mapping): the pipeline assumes the VOMM trained on ATT&CK sequences faithfully reproduces real-world adversary behavior for beam-search path reconstruction and DRL optimization, and that the CSF maturity mapping produces realistic available mitigations with accurate costs. No validation against incident data, cross-checks, sensitivity analysis, or fidelity metrics are reported; deviations here would render the reported policies simulation artifacts rather than evidence of practical utility.

Authors: The referee correctly notes the absence of direct empirical validation for the VOMM fidelity and the CSF-to-mitigation cost mapping. The manuscript (Section 3) describes the data sources, training procedure, and mapping rules, but does not report incident-log validation, sensitivity sweeps, or fidelity metrics. We will add a new subsection that explicitly states the modeling assumptions, discusses potential deviations from real-world behavior, and outlines planned validation steps (including fidelity metrics and sensitivity analysis on cost assignments). This addition will clarify the current scope as a framework demonstration while transparently addressing the practical-utility concern. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation is simulation-driven from external data and frameworks

full rationale

The paper trains a VOMM on observed ATT&CK sequences (external data), maps CSF maturity to mitigations via the NIST framework, reconstructs paths with beam search, and optimizes via DRL under budget constraints. The reported stable policies and cost-risk trade-offs emerge from running the RL agent in this constructed environment; they are not equivalent to the inputs by definition, nor do they rely on fitted parameters renamed as predictions or self-citation chains for uniqueness. No load-bearing step reduces to a tautology or prior author result invoked as an external theorem. This is a standard simulation-optimization pipeline whose validity hinges on external fidelity assumptions rather than internal circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Review performed on abstract only; full text unavailable so ledger is necessarily incomplete. Key unverified assumptions include accuracy of CSF-ATT&CK mapping and fidelity of VOMM to real adversaries.

axioms (2)

domain assumption CSF maturity assessments can be reliably mapped to MITRE ATT&CK mitigation capabilities for planning purposes
Invoked to enable integration of organizational security posture with adversary-informed defensive planning.
domain assumption VOMM trained on observed ATT&CK sequences produces realistic attack-path distributions
Required for scalable adversary simulation inside the DRL environment.

pith-pipeline@v0.9.0 · 5506 in / 1361 out tokens · 46914 ms · 2026-05-12T02:32:02.115622+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We employ a Variable-Order Markov Model (VOMM) trained on observed ATT&CK technique sequences... LLM-assisted semantic translation layer that converts NIST CSF practice assessments into mitigation maturity levels
IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We reconstruct likely attack paths and defensive responses using beam search, and then jointly optimize mitigation selection under explicit budget constraints

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

31 extracted references · 31 canonical work pages

[1]

Mitre att&ck applications in cybersecurity and the way forward.arXiv preprint arXiv:2502.10825, 2025

Yuning Jiang, Qiaoran Meng, Feiyang Shang, Nay Oo, Le Thi Hong Minh, Hoon Wei Lim, and Biplab Sikdar. Mitre att&ck applications in cybersecurity and the way forward.arXiv preprint arXiv:2502.10825, 2025

work page arXiv 2025
[2]

The never-evolving threat landscape: Forever techniques and the illusion of change

Brian Donohue. The never-evolving threat landscape: Forever techniques and the illusion of change. Presentation at ATT&CKcon 6.0, October 2025. MITRE ATT&CK Conference, slides

work page 2025
[3]

Cyborg: A gym for the development of autonomous cyber agents.arXiv preprint arXiv:2108.09118, 2021

Maxwell Standen, Martin Lucas, David Bowman, Toby J Richer, Junae Kim, and Damian Marriott. Cyborg: A gym for the development of autonomous cyber agents.arXiv preprint arXiv:2108.09118, 2021

work page arXiv 2021
[4]

Cage challenge 4: A scalable multi-agent reinforcement learning gym for autonomous cyber defence.AI Magazine, 46(3):e70021, 2025

Mitchell Kiely, Metin Ahiskali, Etienne Borde, Benjamin Bowman, David Bowman, Dirk Van Bruggen, KC Cowan, Prithviraj Dasgupta, Erich Devendorf, Ben Edwards, et al. Cage challenge 4: A scalable multi-agent reinforcement learning gym for autonomous cyber defence.AI Magazine, 46(3):e70021, 2025. 22 DRL-Based Cyber Mitigation Planning Algorithm 1Attack Path R...

work page 2025
[5]

Mitre att&ck: Design and philosophy.MITRE Technical Report, 2018

Blake Strom, Andy Applebaum, Doug Miller, Kathryn Nickels, Adam Pennington, and Cody Thomas. Mitre att&ck: Design and philosophy.MITRE Technical Report, 2018

work page 2018
[6]

Mitre att&ck for industrial control systems, 2024

MITRE Corporation. Mitre att&ck for industrial control systems, 2024. https://attack.mitre.org/matrices/ics/

work page 2024
[7]

Attack flow v3, 2025

The Center for Threat-Informed Defense. Attack flow v3, 2025. https://ctid.mitre.org/projects/attack-flow/

work page 2025
[8]

Executive order 13636: Improving critical infrastructure cybersecurity

Executive Office of the President. Executive order 13636: Improving critical infrastructure cybersecurity. https://www.federalregister.gov/documents/2013/02/19/2013-03915/ improving-critical-infrastructure-cybersecurity, 2013

work page 2013
[9]

Cybersecurity framework (csf) 2.0

National Institute of Standards and Technology. Cybersecurity framework (csf) 2.0. https://www.nist.gov/ cyberframework, 2024

work page 2024
[10]

The finance of cybersecurity: Quantitative modeling of investment decisions and net present value.International Journal of Production Economics, 279:109448, 2025

Mazen Brho, Amer Jazairy, and Aaron V Glassburner. The finance of cybersecurity: Quantitative modeling of investment decisions and net present value.International Journal of Production Economics, 279:109448, 2025

work page 2025
[11]

Optimal network security hardening using attack graph games

Karel Durkota, Viliam Lis `y, Branislav Bosansk `y, and Christopher Kiekintveld. Optimal network security hardening using attack graph games. InIJCAI, pages 526–532, 2015

work page 2015
[12]

Intelligent, automated red team emulation

Andy Applebaum, Doug Miller, Blake Strom, Chris Korban, and Ross Wolf. Intelligent, automated red team emulation. InProceedings of the 32nd annual conference on computer security applications, pages 363–373, 2016

work page 2016
[13]

Heuristic search value iteration for one-sided partially observable stochastic games

Karel Horák, Branislav Bošansk`y, and Michal Pˇechouˇcek. Heuristic search value iteration for one-sided partially observable stochastic games. InProceedings of the AAAI conference on artificial intelligence, volume 31, 2017

work page 2017
[14]

A Game-Theoretical Approach to Cyber-Security of Critical Infrastructures Based on Multi-Agent Reinforcement Learning, 2018

work page 2018
[15]

Causally aware reinforcement learning agents for autonomous cyber defence.Knowledge-Based Systems, 304:112521, 2024

Tom Purves, Konstantinos G Kyriakopoulos, Sian Jenkins, Iain Phillips, and Tim Dudman. Causally aware reinforcement learning agents for autonomous cyber defence.Knowledge-Based Systems, 304:112521, 2024

work page 2024
[16]

A comprehensive survey: Evaluating the efficiency of artificial intelligence and machine learning techniques on cyber security solutions.IEEe Access, 12:12229–12256, 2024

Merve Ozkan-Okay, Erdal Akin, Ömer Aslan, Selahattin Kosunalp, Teodor Iliev, Ivaylo Stoyanov, and Ivan Beloev. A comprehensive survey: Evaluating the efficiency of artificial intelligence and machine learning techniques on cyber security solutions.IEEe Access, 12:12229–12256, 2024

work page 2024
[17]

Large language models are autonomous cyber defenders.arXiv [cs.AI], July 2025

Sebastián R Castro, Roberto Campbell, Nancy Lau, Octavio Villalobos, Jiaqi Duan, and Alvaro A Cardenas. Large language models are autonomous cyber defenders.arXiv [cs.AI], July 2025. 23 DRL-Based Cyber Mitigation Planning

work page 2025
[18]

Design and evaluation of an autonomous cyber defence agent using DRL and an augmented LLM.Comput

Johannes Loevenich, Erik Adler, Tobias Hürten, and Roberto Rigolin F Lopes. Design and evaluation of an autonomous cyber defence agent using DRL and an augmented LLM.Comput. Netw., 262(111162):111162, May 2025

work page 2025
[19]

Large language model-based reward design for deep reinforcement learning-driven autonomous cyber defense.arXiv [cs.LG], November 2025

Sayak Mukherjee, Samrat Chatterjee, Emilie Purvine, Ted Fujimoto, and Tegan Emerson. Large language model-based reward design for deep reinforcement learning-driven autonomous cyber defense.arXiv [cs.LG], November 2025

work page 2025
[20]

The path to autonomous cyberdefense.IEEE Secur

Sean Oesch, Phillipe Austria, Amul Chaulagain, Brian Weber, Cory Watson, Matthew Dixson, and Amir Sadovnik. The path to autonomous cyberdefense.IEEE Secur. Priv., 23(1):38–46, January 2025

work page 2025
[21]

Entity-based reinforcement learning for autonomous cyber defence

Isaac Symes Thompson, Alberto Caron, Chris Hicks, and Vasilios Mavroudis. Entity-based reinforcement learning for autonomous cyber defence. InProceedings of the Workshop on Autonomous Cybersecurity, pages 56–67, 2023

work page 2023
[22]

{AIRS}: Explanation for deep reinforcement learning based security applications

Jiahao Yu, Wenbo Guo, Qi Qin, Gang Wang, Ting Wang, and Xinyu Xing. {AIRS}: Explanation for deep reinforcement learning based security applications. In32nd USENIX Security Symposium (USENIX Security 23), pages 7375–7392, 2023

work page 2023
[23]

Adaptive cyber defense using advanced deep reinforcement learning algorithms: a real-time comparative analysis.Journal of Computing Theories and Applications, 2(4):523–535, 2025

Atheer Alaa Hammad and Firas Tarik Jasim. Adaptive cyber defense using advanced deep reinforcement learning algorithms: a real-time comparative analysis.Journal of Computing Theories and Applications, 2(4):523–535, 2025

work page 2025
[24]

Multi-agent reinforcement learning for maritime operational technology cyber security.arXiv preprint arXiv:2401.10149, 2024

Alec Wilson, Ryan Menzies, Neela Morarji, David Foster, Marco Casassa Mont, Esin Turkbeyler, and Lisa Gralewski. Multi-agent reinforcement learning for maritime operational technology cyber security.arXiv preprint arXiv:2401.10149, 2024

work page arXiv 2024
[25]

Probabilistic attack sequence generation and execution based on mitre att&ck for ics datasets

Seungoh Choi, Jeong-Han Yun, and Byung-Gil Min. Probabilistic attack sequence generation and execution based on mitre att&ck for ics datasets. InProceedings of the 14th Cyber Security Experimentation and Test Workshop, pages 41–48, 2021

work page 2021
[26]

Mitre att&ck-driven cyber risk assessment

Mohamed Ahmed, Sakshyam Panda, Christos Xenakis, and Emmanouil Panaousis. Mitre att&ck-driven cyber risk assessment. InProceedings of the 17th International Conference on Availability, Reliability and Security, pages 1–10, 2022

work page 2022
[27]

Att&ck behavior forecasting based on collaborative filtering and graph databases

Masaki Kuwano, Momoka Okuma, Satoshi Okada, and Takuho Mitsunaga. Att&ck behavior forecasting based on collaborative filtering and graph databases. In2022 IEEE International Conference on Computing (ICOCO), pages 191–197. IEEE, 2022

work page 2022
[28]

Linking threat agents to targeted organizations: A pipeline for enhanced cybersecurity risk metrics

Spencer Massengale and Philip Huff. Linking threat agents to targeted organizations: A pipeline for enhanced cybersecurity risk metrics. In2024 4th Intelligent Cybersecurity Conference (ICSC), pages 132–141. IEEE, 2024

work page 2024
[29]

Large language models for cyber threat intelligence: Extracting mitre with llms

Andraž Krašovec, Gary Steri, Georgios Karopoulos, and Mirko Trapani. Large language models for cyber threat intelligence: Extracting mitre with llms. InInternational Conference on Availability, Reliability and Security, pages 80–89. Springer, 2025

work page 2025
[30]

Staying ahead of threat actors in the age of AI

Microsoft Security and OpenAI. Staying ahead of threat actors in the age of AI. https://www.microsoft. com/en-us/security/blog/2024/02/14/staying-ahead-of-threat-actors-in-the-age-of-ai/ , February 2024. Industry threat intelligence report

work page 2024
[31]

Assessing and prioritizing ransomware risk based on historical victim data

Spencer Massengale and Philip Huff. Assessing and prioritizing ransomware risk based on historical victim data. InInternational Conference on Security and Privacy in Communication Systems, pages 351–369. Springer, 2024. 24

work page 2024