A multi-stage coordinated cyber-physical topology attack method based on deep reinforcement learning
CSTR:
Author:
Affiliation:

Clc Number:

TM732

Fund Project:

  • Article
  • |
  • Figures
  • |
  • Metrics
  • |
  • Reference
  • |
  • Related
  • |
  • Cited by
  • |
  • Materials
  • |
  • Comments
    Abstract:

    With the development of smart grid and the continuous introduction of communication equipments into cyber physical system (CPS),CPS is confronted with a new attack mode with more destructive—coordinated cyber physical attack (CCPA). CCPA is not only hidden but also threatening,which is easy to cause cascading failures. Firstly,from the perspective of the attacker,a multi-stage coordinated cyber-physical topology attack model is proposed. The single-stage physical attack first trips a transmission line,and the two-stage cyber attack is used to mask the outage signal of the disconnected line in the physical layer and then create a new fake tripped line in the cyber layer. Secondly,combined with deep reinforcement learning (DRL) theory,the method for determining the minimum attack resources based on deep Q-network (DQN) is proposed. Then,the specific model and solution method for the attacker are given,taking the maximization of the physical attack effect in the upper layer and minimization of the attack cost in the lower layer into consideration. Finally,the IEEE 30-bus system is taken as an example to verify the effectiveness of the proposed multi-stage attack model. The simulation results demonstrate that the multi-stage coordinated cyber-physical topology attack is more hidden and effective than the single attack,and the damage to the power grid is greater,which provides a reference for the defender against such attacks.

    Reference
    Related
    Cited by
Get Citation
Share
Article Metrics
  • Abstract:
  • PDF:
  • HTML:
  • Cited by:
History
  • Received:January 18,2023
  • Revised:March 09,2023
  • Adopted:March 10,2023
  • Online: July 20,2023
  • Published: July 28,2023
Article QR Code