Objective The essence of cyberspace security lies in attack-defense games under uncertainty, and the outcome of such games depends on the capability to characterize and control uncertainty. As cyberspace confrontation has extended from the network information domain to the physical domain, electromagnetic spectrum domain, and social cognition domain, uncertainty has increased significantly. The coupling among heterogeneous domains introduces new uncertainty sources and leads to cross-domain propagation and cascading amplification of uncertainty, making confrontation modeling and decision-making increasingly challenging. Traditional single-domain confrontation models are insufficient for characterizing such cross-domain uncertainty propagation. However, existing studies lack a unified uncertainty quantification framework that supports integrated modeling, reasoning, and decision-making for multi-domain fusion confrontation. This paper aimed to establish an entropy-evolution-driven intelligent cybersecurity paradigm, in which multi-domain fusion attack-defense games were characterized as dynamic entropy control processes. Methods Within this framework, entropy was introduced as a unified mathematical metric for quantifying heterogeneous uncertainty across multiple domains, and entropy evolution was employed to characterize the dynamic variation of uncertainty states driven by attack-defense interactions. The attack and defense processes were modeled as bidirectional entropy control mechanisms, where both sides attempted to regulate entropy evolution toward favorable states by reducing their own entropy while increasing the opponent entropy. To characterize the uncertainty states underlying entropy evolution, the attack-defense game was modeled from a probabilistic perspective, where the cognitive states of both sides were represented by belief distributions over uncertain environments and opponent strategies. A dual entropy metric framework integrating attack entropy and defense entropy was constructed, where Shannon entropy and relative entropy were incorporated to quantify cognitive uncertainty and characterize attack-defense strategy effectiveness within a unified entropy-based metric framework. The resulting entropy representation provided a common state space for describing the uncertainty evolution of both attack and defense processes, enabling subsequent dynamic analysis and strategic reasoning. Based on the entropy metrics, a cross-domain uncertainty dynamics model was constructed to describe the propagation and amplification of uncertainty among heterogeneous domains under attack-defense interactions. A stochastic difference equation was formulated to characterize entropy evolution, where cross-domain coupling effects were incorporated to capture the cascading transmission of uncertainty. A formal criterion for attack-defense strategy effectiveness was further established through Lyapunov stability theory to evaluate the convergence and stability of entropy evolution under different strategies. This criterion provided a theoretical basis for distinguishing effective strategies from ineffective strategies according to their influence on entropy evolution trajectories. The entropy control logic was further embedded into a game-theoretic reasoning framework to support adaptive attack-defense strategy generation. A hierarchical game-theoretic reasoning model was employed as the underlying reasoning environment, integrating a macroscopic stage-transition model based on Markov decision processes and a microscopic behavior reasoning model based on dynamic Bayesian networks. Based on the dual entropy representation, a belief-driven multi-agent strategy generation method with symmetric opponent modeling was incorporated to characterize both-side cognitive states and generate adaptive strategies under uncertain conditions, forming a closed-loop process among belief updating, entropy evolution, and strategy generation. Through this mechanism, uncertainty quantification, game-theoretic reasoning, and adaptive strategy generation were integrated into a unified computational framework. Results and Discussions Numerical reasoning simulations were conducted in a multi-domain security protection scenario of a regional power grid control center to validate the proposed entropy-evolution-driven framework. The results showed that attack entropy and defense entropy transformed heterogeneous uncertainties from the physical, electromagnetic spectrum, network information, and social cognition domains into comparable entropy representations. The analysis of cross-domain coupling effects demonstrated that ignoring uncertainty propagation among domains could lead to pseudo-convergence and misjudgment of strategy effectiveness. Compared with the uncoupled reasoning approach (Baseline A), the proposed method effectively captured the cascading amplification of uncertainty caused by cross-domain interactions and avoided the misjudgment of locally effective but globally ineffective strategies. Compared with the conventional single-domain threshold-based alerting approach (Baseline B), the proposed method enabled more effective defense decisions by integrating multi-domain information fusion, entropy-control objectives, and symmetric opponent modeling. Furthermore, the scenario reasoning demonstrated the closed-loop coupling mechanism between the microscopic-level reasoning process and the macroscopic-level stage transition process in the hierarchical game-theoretic model. The microscopic layer performed belief updating, opponent mirror modeling and reasoning, entropy evolution prediction, and strategy selection within the current confrontation stage, while the resulting entropy evolution provided quantitative criteria for macroscopic-level stage transition. This mechanism enabled the transformation from uncertainty quantification and strategic reasoning to decision-making and stage evolution, forming a unified computational loop for multi-domain fusion confrontation. Conclusions The proposed paradigm establishes an entropy-evolution-driven framework for constructing multi-domain fusion cybersecurity architectures. Under this paradigm, defense evolution from kill-chain-oriented operations to kill-web-oriented architectures was interpreted through entropy control, where defensive effectiveness was achieved through coordinated regulation of uncertainty across multiple domains. By integrating entropy-based uncertainty quantification, dynamic modeling, and game-theoretic reasoning, the framework provides a mathematical basis for transforming cyberspace security decision-making from experience-dependent approaches toward computable paradigms. Future research will focus on improving the parameterization of belief distributions and the tractability of entropy computation, advancing the structural identification and parameter acquisition of cross-domain uncertainty dynamics models, and enhancing the real-time scalability of game-theoretic reasoning with large strategy spaces.
ChoD X, NguyenT T.A novel approach for APT attack detection based on an advanced computing[J].Scientific Reports,2024,14:22223. doi:10.1038/s41598-024-72957-0
[8]
WangZhiwei, HeXijie, YiXin,et al.Survey of attack and detection based on the full life cycle of APT[J].Journal on Communications,2024,45(9):206‒228. doi:10.11959/j.issn.1000-436x.2024128
DuanYancui, CaiYonghua, WangZhikang,et al.A novel network security risk assessment approach by combining subjective and objective weights under uncertainty[J].Applied Sciences,2018,8(3):428. doi:10.3390/app8030428
[11]
WuJiacheng, YuXiao.A review of research on cybersecurity risk assessment methods[J].Electronic Science and Technology,2024,37(3):10‒17.
[12]
吴嘉诚,余晓.网络安全风险评估方法研究综述[J].电子科技,2024,37(3):10‒17.
[13]
WuJunlin, KamhouaC, KantarciogluM,et al.Learning generative deception strategies in combinatorial masking games[C]//Decision and Game Theory for Security.Cham:Springer,2021:98‒117. doi:10.1007/978-3-030-90370-1_6
[14]
ZhangTao, XuChangqiao, ShenJiahao,et al.How to disturb network reconnaissance:A moving target defense approach based on deep reinforcement learning[J].IEEE Transactions on Information Forensics and Security,2023,18:5735‒5748. doi:10.1109/tifs.2023.3314219
[15]
SlusherM N.Lessons from the Ukraine conflict:Modern warfare in the age of autonomy,information,and resilience:[R].Washington D C:Center for Strategic and International Studies(CSIS),2025.
[16]
National Institute for Defense Studies (Japan).Future of Multi-domain Operations and Campaigning:lessons for the Indo-Pacific[R].Tokyo:NIDS,2025.
[17]
HuangMengyu, TsangK F E, LiYuzhe,et al.Strategic DoS attack in continuous space for cyber-physical systems over wireless networks[J].IEEE Transactions on Signal and Information Processing Over Networks,2022,8:421‒432. doi:10.1109/tsipn.2022.3174969
[18]
ChenJun, SunXin, TianWen,et al.A Bayesian game approach for node-based attribution defense against asymmetric information attacks in IoT networks[J].PLoS One,2025,20(3):e0316091. doi:10.1371/journal.pone.0316091
[19]
WuDandan, ChenJie, XieRuiyun,et al.OntoCSD:An ontology-based security model for an integrated solution of cyberspace defense[J].Frontiers of Information Technology & Electronic Engineering,2024,25(9):1209‒1225. doi:10.1631/fitee.2300662
[20]
TangWei, LiuYujun, ChenYuling,et al.SLBRS:Network virus propagation model based on safety entropy[J].Applied Soft Computing,2020,97:106784. doi:10.1016/j.asoc.2020.106784
[21]
LiuKe, WangMufeng, MaRongkuan,et al.Detection and localization of cyber attacks on water treatment systems:An entropy-based approach[J].Frontiers of Information Technology & Electronic Engineering,2022,23(4):587‒603. doi:10.1631/fitee.2000546
SinhaM, MukherjeeJ C, BeraP,et al.SynFloWatch:An entropy-based live defense system against SYN spoofing DDoS attacks in hybrid SDN[J].Journal of Network and Systems Management,2025,33(4):76. doi:10.1007/s10922-025-09934-z
[26]
WangMiao, WangLiming, XuZhen,et al.Research on DDoS detection in multi-tenant cloud based on entropy change[J].Journal on Communications,2016,37(Z1):204‒210.
PaikJ Y, JinRize, ChoE S.Malware classification using a byte-granularity feature based on structural entropy[J].Computational Intelligence,2022,38(4):1536‒1558. doi:10.1111/coin.12521
[29]
SafaeiF, YeganlooH, AkbarR.Robustness on topology reconfiguration of complex networks:An entropic approach[J].Mathematics and Computers in Simulation,2020,170:379‒409. doi:10.1016/j.matcom.2019.11.013
[30]
YuHaiyun, YangXincun, WangLiang.Network security situation prediction based on combining associated entropy and deep recurrent neural network[J].Transactions on Emerging Telecommunications Technologies,2022,33(10):e4164. doi:10.1002/ett.4164
[31]
ZhaoXiaolin, GuoJiong, JiangXiaoyi,et al.Multi-layer network attack and defense model based on entropy method to measure network security[J].Journal of Physics:Conference Series,2019,1302(2):022068. doi:10.1088/1742-6596/1302/2/022068
JananiK.Cybersecurity through entropy injection:A paradigm shift from reactive defense to proactive uncertainty[PP/OL].(2025‒04‒15)[2026‒05‒01].
[34]
WuJiangxing.Development paradigms of cyberspace endogenous safety and security[J].Scientia Sinica(Informationis),2022,52(2):189‒204.
[35]
邬江兴.网络空间内生安全发展范式[J].中国科学:信息科学,2022,52(2):189‒204.
[36]
WuJiangxing.Research on cyber mimic defense[J].Journal of Cyber Security,2016,1(4):1‒10.
[37]
邬江兴.网络空间拟态防御研究[J].信息安全学报,2016,1(4):1‒10.
[38]
MaHailong, WangLiang, HuTao,et al.Survey on the development of mimic defense in cyberspace:From mimic concept to "mimic+"ecology[J].Chinese Journal of Network and Information Security,2022,8(2):15‒38.
SontagE D.Mathematical control theory:Deterministic finite dimensional systems[M].2nd ed.New York:Springer,1998.
[43]
BerkenkampF, TurchettaM, SchoelligA P,et al.Safe model-based reinforcement learning with stability guarantees[C]//Proceedings of the Advances in Neural Information Processing Systems 30 (NeurIPS).Long Beach:Curran Associates,2017:6692.
[44]
WangQian, WangWeilong, WangYan,et al.Multi-stage network attack detection algorithm based on Gaussian mixture hidden Markov model and transfer learning[J].IEEE Transactions on Automation Science and Engineering,2025,22:3470‒3484. doi:10.1109/tase.2024.3395355
[45]
BandaT V, RensG B.Reinforcement learning of defensive strategies against attacks in partially observable local area networks[J].Advances in Knowledge-Based Systems,Data Science,and Cybersecurity,2025,2(3):337‒374.
[46]
HutchinsE M, CloppertM J, AminR M.Intelligence-driven computer network defense informed by analysis of adversary campaigns and intrusion kill chains[C]//Proceedings of the Leading Issues in Information Warfare & Security Research.2011,1(1):80.
[47]
YangYating, ZhuQuanyan.Toward a multi-echelon cyber warfare theory:A meta-game-theoretic paradigm for defense and dominance[EB/OL].(2026‒02‒24)[2026‒05‒01].