Neurotechnology is safe
A fragment of the report “Neurotechnologies in security” at the ZeroNights'16 conference on the use of brain activity data in biometric control systems.
When using data on brain activity as a biometric parameter, the overall architecture of the verification system is preserved.
An authorized user has an identifier and a biometric sample. For verification, data from the database and from the user are compared. But there are subtleties in working with brain activity data.
We obtain data on brain activity by electroencephalography .
EEG captures the total electrical activity of the brain by the potential difference on the surface of the scalp. Potentials are removed by special electrodes.
The EEG signal is non-linear, noisy and non-stationary.
Removal of the EEG - contact procedure: you need to install the electrodes on the user's head. And long, because the electrical activity of the brain is extended in time. We need to collect a certain amount of data, while they cannot be taken for a long time. Over long periods of time, nonlinear distortions of the EEG signal appear.
We cope with non-linearity with a series of short measurements, during which the signal can be considered linear.
EEG may contain signals that are not related to brain activity. These are artifacts .
Artifacts may be of a physical or physiological nature.
EEG is affected by muscle contraction, and interference from electrical appliances, and the quality of electrodes.
You can increase the purity of the signal procedurally, hardware, software.
It is best to shoot an eeg in a relaxed atmosphere when the user is not moving, relaxed and concentrated. External stimuli that repeat from procedure to procedure can be added. For example, the same music, pictures.
Artifacts can be recognized in person. It is necessary to remove the EEG when a person blinks, nods, clenches his jaw, smiles, speaks. Artifacts will appear on the electroencephalogram. You can supplement the EEG with myo sensors and accelerometers that record muscle contraction and head movement. Then cut out sections with extraneous signals from the electroencephalogram.
The quality of the EEG signal directly depends on the quality of the contact of the electrodes with the surface of the scalp. It is important to correctly position the electrodes and reduce the resistance between them and the skin, for this you can use conductive gels or saline.
The recorded signal must be cleaned of noise. For example, using filter systems based on band-pass . To improve the purity of the signal, you can fine-tune the passband of the filter. The bandwidth depends on the particular neural interface.
After cleaning the signal, you need to highlight its significant features. This process is called - feature extraction .
Feature extraction is the acquisition of the characteristics of the most informative signal fragments. The obtained characteristics can be used in classification problems.
For EEG processing, you can use Fast Fourier Transform , as a result we obtain the frequency characteristics of the signal.
However, the FFT is a linear method, and the EEG signal is non-stationary.
For processing an unsteady signal, time-frequency analysis methods are more suitable. For example, a wavelet transform .
The wavelet transform represents an EEG signal as a sequence of wavelets . This allows us to consider the frequency component of the EEG in the time perspective and provides a clear reference of the spectrum of significant signal attributes to time.
The last stage of working with EEG in the verification system is a biometric matcher.
For all its limitations, EEG has the potential for use in biometric systems.
EEG can be used as a biometric parameter , because brain activity is individual . It is unique in the synchronized activity of groups of neurons.
Neurons that process the same signals form metastable groups.
The signals corresponding to one external stimulus or cognitive event cause synchronized activity of neurons united into groups. A certain level of such synchronization is maintained at rest.
Synchronized activity of neurons is observed on electroencephalograms.
EEG as a biometric parameter has advantages :
When using data on brain activity as a biometric parameter, the overall architecture of the verification system is preserved.
An authorized user has an identifier and a biometric sample. For verification, data from the database and from the user are compared. But there are subtleties in working with brain activity data.
We obtain data on brain activity by electroencephalography .
EEG captures the total electrical activity of the brain by the potential difference on the surface of the scalp. Potentials are removed by special electrodes.
The EEG signal is non-linear, noisy and non-stationary.
Removal of the EEG - contact procedure: you need to install the electrodes on the user's head. And long, because the electrical activity of the brain is extended in time. We need to collect a certain amount of data, while they cannot be taken for a long time. Over long periods of time, nonlinear distortions of the EEG signal appear.
We cope with non-linearity with a series of short measurements, during which the signal can be considered linear.
EEG may contain signals that are not related to brain activity. These are artifacts .
Artifacts may be of a physical or physiological nature.
EEG is affected by muscle contraction, and interference from electrical appliances, and the quality of electrodes.
You can increase the purity of the signal procedurally, hardware, software.
It is best to shoot an eeg in a relaxed atmosphere when the user is not moving, relaxed and concentrated. External stimuli that repeat from procedure to procedure can be added. For example, the same music, pictures.
Artifacts can be recognized in person. It is necessary to remove the EEG when a person blinks, nods, clenches his jaw, smiles, speaks. Artifacts will appear on the electroencephalogram. You can supplement the EEG with myo sensors and accelerometers that record muscle contraction and head movement. Then cut out sections with extraneous signals from the electroencephalogram.
The quality of the EEG signal directly depends on the quality of the contact of the electrodes with the surface of the scalp. It is important to correctly position the electrodes and reduce the resistance between them and the skin, for this you can use conductive gels or saline.
The recorded signal must be cleaned of noise. For example, using filter systems based on band-pass . To improve the purity of the signal, you can fine-tune the passband of the filter. The bandwidth depends on the particular neural interface.
After cleaning the signal, you need to highlight its significant features. This process is called - feature extraction .
Feature extraction is the acquisition of the characteristics of the most informative signal fragments. The obtained characteristics can be used in classification problems.
For EEG processing, you can use Fast Fourier Transform , as a result we obtain the frequency characteristics of the signal.
However, the FFT is a linear method, and the EEG signal is non-stationary.
For processing an unsteady signal, time-frequency analysis methods are more suitable. For example, a wavelet transform .
The wavelet transform represents an EEG signal as a sequence of wavelets . This allows us to consider the frequency component of the EEG in the time perspective and provides a clear reference of the spectrum of significant signal attributes to time.
The last stage of working with EEG in the verification system is a biometric matcher.
For all its limitations, EEG has the potential for use in biometric systems.
EEG can be used as a biometric parameter , because brain activity is individual . It is unique in the synchronized activity of groups of neurons.
Neurons that process the same signals form metastable groups.
The signals corresponding to one external stimulus or cognitive event cause synchronized activity of neurons united into groups. A certain level of such synchronization is maintained at rest.
Synchronized activity of neurons is observed on electroencephalograms.
EEG as a biometric parameter has advantages :
- It can only be removed explicitly with the knowledge and consent of the user, because the procedure is contact and requires concentration
- EEG can be used for identity verification. If someone replaces an authorized user at work in the system, then this can be tracked by a change in the nature of brain activity.
- The electroencephalogram is variable. To create a new biometric sample, it is enough to concentrate on something else. Theoretically, we can create an unlimited number of biometric samples on one physiological property.
- EEG reflects the psycho-emotional state of the user. EEG is a physiological manifestation of behavioral characteristics.