A novel power system APT attack graph generation method based on GD-DQN algorithm

By combining the GD-DQN algorithm with Dueling DQN and the GRU model, the problems of scale and redundancy in attack graph generation in new power systems are solved, achieving efficient, dynamic, and real-time attack graph generation and optimized path selection, thereby improving the defense capabilities of power systems.

CN115271029BActive Publication Date: 2026-06-26NORTH CHINA ELECTRIC POWER UNIV +1
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
NORTH CHINA ELECTRIC POWER UNIV
Filing Date
2022-06-10
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Traditional attack graph generation methods suffer from node and edge explosion problems in large-scale new power systems, resulting in large and redundant attack graphs that affect practicality. At the same time, Q-learning algorithms are difficult to store in high-dimensional continuous state spaces, have a single reward function, and existing methods are subjective in assessing vulnerability risks.

Method used

The GD-DQN algorithm is adopted, combined with Dueling DQN and GRU models. Network awareness is generated through network vulnerability scanning. The agent performs dynamic real-time attack graph generation in the power system. The attack path is optimized by using a memory bank and an improved reward function. The GRU model is integrated to enhance the agent's perception and memory functions.

Benefits of technology

It enables efficient and dynamic real-time generation of attack graphs in new power systems, improves the scale and quality of attack graphs, enhances the flexibility and robustness of defense systems, and optimizes the selection of attack paths.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115271029B_ABST
    Figure CN115271029B_ABST
Patent Text Reader

Abstract

The application relates to a novel power system APT attack graph generation method based on a GD-DQN algorithm, first, based on a novel power system network topology structure, network vulnerability scanners are used for scanning detection to generate vulnerability information of the novel power system; second, an intelligent agent in the GD-DQN algorithm is used for carrying out vulnerability state scanning on the vulnerability information to form network perception T of the novel power system environment; and then, the intelligent agent constructs an attack graph through a training process according to the network perception T. In the application, the intelligent agent in the GD-DQN algorithm is used to improve the efficiency and scale of attack graph generation, and the attack graph can be dynamically and real-timely generated; the Dueling DQN algorithm is introduced, each learning process of the intelligent agent does not necessarily depend on other learning scenes; the GRU model is introduced, the intelligent agent achieves better training effect, and a better attack path can be obtained.
Need to check novelty before this filing date? Find Prior Art