A self-evolution cyber attack scheme generation system for cybersecurity evaluation
Mingsheng Yang, Yan Jia, Yangyang Mei, Jie Yang, Weihong Han, Jiawei Zhang, Zhuocheng Yu · Scientific Reports · 2026
With the increasing complexity of network attacks, defense systems face significant challenges in maintaining cybersecurity. To effectively evaluate and optimize defense strategies, this paper proposes a self-evolution attack scenario generation system tailored for assessment purposes. To address the scalability challenges in attack graph generation and improve the efficiency and relevance of security evaluations, the system incorporates a real-time generation method capable of dynamically adapting attack scenarios based on specific goals and constraints. Additionally, a methodology is developed to construct potential attack paths using attack graph techniques enhanced with self-evolving mechanisms. The feasibility and adaptability of the generated attack scenarios are validated through simulation experiments. This paper details the system's design, highlighting its core technical innovations-including incremental graph updates, scalable goal-driven path generation, and quantitative path ranking-which address key limitations of traditional tools like MulVAL. The system's effectiveness and superiority in scalability and usability are demonstrated through extensive simulations.