Cognitive radio space-frequency two-dimensional anti-hostility jamming method based on deep reinforcement learning

Abstract

The invention discloses a cognitive radio space-frequency two-dimensional anti-hostility jamming method based on deep reinforcement learning. A cognitive radio secondary user observes an access state of a cognitive radio primary user and a signal to jamming ratio of a wireless signal under a state of unknowing a jammer attack mode and a wireless channel environment, and decides whether to leave the located interfered region or select an appropriate frequency point to send the signal by use of a deep reinforcement learning mechanism. A deep convolutional nerve network and Q learning are combined, the Q learning is used for learning an optimal anti-jamming strategy in a wireless dynamic game, and an observation state and acquired benefit are input into the deep convolutional nerve network as a training set to accelerate the learning speed. By use of the deep reinforcement learning mechanism, the communication efficiency for competing hostility jammer by the cognitive radio under a wireless network environment scene in dynamic change is improved. A problem that the learning speed is fast reduced since an artificial nerve network needs to firstly classify the data in the training process and the Q learning algorithm is large in dimension in a state set and an action set can be overcome.

Description

technical field [0001] The invention relates to wireless network security, in particular to a cognitive radio space-frequency two-dimensional anti-hostile interference method based on deep reinforcement learning. Background technique [0002] With the rapid development of wireless communication, problems such as the shortage and utilization of spectrum resources are becoming more and more serious. The proposal of cognitive radio (Cognitive Radio, CR) technology can effectively improve the utilization of spectrum. Since cognitive radio adopts an open spectrum and a dynamic access method, it is extremely vulnerable to hostile interference attacks, and its security issues need to be resolved urgently. [0003] The hostile jammer prevents legitimate users from forwarding normal data by occupying the communication channel of network nodes, and then launches a denial of service attack (DoS). As a traditional anti-interference technology, spread spectrum communication can effectiv...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a recognition method based on deep reinforcement learning that can overcome the problem that the artificial neural network needs to classify the data first in the training process and that the learning speed of the Q learning algorithm will decrease rapidly when the state set and action set dimensions are large. Two-dimensional anti-hostile jamming method for radio space frequency

Images