An attack method against EEG brain-computer interface based on convolution neural network

Abstract

The invention discloses an attack method for EEG brain computer interface based on convolution neural network, includes using white box attack to construct white box countermeasure sample of EEG brainsignal, using grey box attack to construct grey box countermeasure sample of EEG brain signal, using black box attack to construct black box countermeasure sample of EEG brain signal, using white boxattack to construct white box countermeasure sample of EEG brain signal, using grey box attack to construct grey box countermeasure sample of EEG brain signal. EEG brain signal white box countermeasure samples, gray box countermeasure samples, black box countermeasure samples attack EEG brain interface convolution neural network. The invention provides three attack methods according to differentattack situations, None of the three attacks need to know the truth label of EEG samples in advance, It is more suitable for the application scenario of BCI, which makes up for the blank of lack of security test of BCI against countermeasure samples, and ensures its effectiveness and practicability, so it can be a good method to test the robustness of EEG BCI system.

Description

technical field [0001] The invention belongs to the technical field of brain-computer interface security, and more specifically relates to an attack method for an EEG brain-computer interface based on a convolutional neural network. Background technique [0002] A brain-computer interface is a system in which the human brain communicates directly with the outside world (computer or other external devices). The electroencephalogram (electroencephalogram, EEG) brain-computer interface system refers to collecting the EEG signals of the human cerebral cortex through the electrodes on the EEG cap, and then analyzing the EEG signals to determine the state or idea of ​​the brain, and then through A system in which EEG signals control external devices. Convolutional neural network is a basic model in deep learning, which has been widely used in image, speech, natural language processing and other fields in recent years. In view of the fact that the convolutional neural network doe...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the defects of the prior art, the purpose of the present invention is to solve the technical problem that there is no better attack method in the current EEG brain-computer interface system to test the robustness of its internal convolutional neural network under different circumstances

Images