Content uploaded by Mohamed Roushdy
Author content
All content in this area was uploaded by Mohamed Roushdy
Content may be subject to copyright.
The 8th International Conference on INFOrmatics and Systems (INFOS2012) – 14-16 May
Bio-inspired Optimization Algorithms and Their Applications Track
Faculty of Computers and Information - Cairo University BIO-55
A Survey of EEG Based User Authentication
Schemes
W. Khalifa , A. Salem, M. Roushdy
Computer Science Department,
Faculty of Computer and Information Sciences,
Ain Shams University,
Cairo, Egypt
abmsalem@yahoo.com
K. Revett
Faculty of Informatics and Computer Science
The British University in Cairo,
Cairo, Egypt
Ken.revett@bue.edu.eg
Abstract— Electroencephalography (EEG) is the recording of
electrical activity occurring in the brain, which is recorded from
the scalp through placement of voltage sensitive electrodes. It has
been repeatedly demonstrated that the brain emits voltage
fluctuations on a continuous basis. These fluctuations are a
reflection of the on-going brain dynamics, which present as a
series of fluctuations that have characteristic waveforms and
amplitude patterns, depending on the cognitive state of the
subject. A number of published reports have indicated that there
is enough depth in the EEG recording, rendering it suitable as a
tool for person authentication. This idea has a solid
underpinning in that recent evidence suggests much of the on-
going EEG recordable activity within brains has a genetic
component. This study presents the common steps for developing
a human identification systems based on EEG signals. It will also
present some of the important techniques used.
Keywords-component; artificial intelligence; behavioral
biometrics; cognitive biometrics; EEG; user identification;signal
processing
I. INTRODUCTION
Biometrics is the process of uniquely identifying
individuals based on one or more physical or behavioral
characteristics. Physiological biometrics is related to the shape
of the body such as finder print, face, and DNA; while
behavioral biometrics is related to the person’s behavior such
as typing rhythm, gait and signature. There are several
techniques of recording brain activities such as
magnetoencephalography (MEG) and functional magnetic
resonance imaging (fMRI), electroencephalography and
Electroencephalography (EEG). EEG signals are brain
activities recorded from electrodes mounted on the scalp. EEG
is the most practical capturing method that can be used in
biometrics due to the advances in its hardware devices; there
are some a EEG signal capturing device that are equal in size to
a mobile phone or computer headset. EEG is one of the
physiological unique characteristics of an individual [1].
There are several requirements that need to be covered by
any biometric system: [2]
a) Changeability: If the user’s authentication information is
compromised, we must be able to replace this information (and
revoke any old password or access credential).
b) Shoulder-surfing resistanc: The scheme must not be
vulnerable to shoulder-surfing, particularly in the presence of
ubiquitous visual recording devices.
c) Theft protection: This includes physical theft and the
computational infeasibility of guessing attacks. If we must rely
on the entropy of an authentication scheme for protection
against off-line dictionary attack, we require an authentication
method whose entropy can scale with processor speeds
d) Protection from user non-compliance: To discourage
unintended transfer to other parties, the user should not be able
to write down (in a manner useful to an attacker) or share their
authentication information “too easily”.
The discovery of electrical currents in the brain was
discovered in the 19th century, but understanding the meaning
of such currents advanced in the past years more rapidly as the
technology improved allowing researchers to capture more data
accurately. Moreover, the advancement in the signal processing
and data classification techniques helped researches to use the
data captured in disease diagnosis, brain computer interface
and finally user identification. Several classifiers were
experimented for the use of EEG signals in user identification
such as neural networks, fisher’s classifier and linear classifier.
In Each experiment the subjects were asked to do one or more
mental tasks such as solving a mathematical problem, mental
visual counting, composing a letter or even just resting.
Depending on the tasks and the classifiers used several
experiments were conducted that showed very promising
results in the use of this characteristic as a behavioral biometric
feature.
The need of a new behavioral biometric is derived from the
need of securing important facilities and important information.
Most of the market available secure systems can be penetrated
by hacking or by a mistake of one the authorized personnel.
The good thing about using EEG is that it covers all the above
The 8th International Conference on INFOrmatics and Systems (INFOS2012) – 14-16 May
Bio-inspired Optimization Algorithms and Their Applications Track
Faculty of Computers and Information - Cairo University BIO-56
mentioned requirements unlike other techniques. Users can
change their password by selecting a different mental tasks, its
prone to shoulder surfing no one can view your thoughts, Users
have to do the authentication their selves, they can’t give a
copy of the password and they have to be alive. One of the
most common used authentication techniques is finger print
recognition; if the user fingerprint is captured by an intruder,
when the system administrators discover this breach the user
and the intruder will be prevented from entering the system,
The benefits of EEG based authentication system is enormous,
it have all the benefits of different authentication methods.
In this paper we investigate the various techniques and
experiments that were developed for using EEG signals as a
user identification characteristic. In the next section we present
a brief summary on the medical aspects of the EEG signals In
section three we describe the methods of capturing EEG signals
and user identification methodology. In section four we discuss
the techniques used to identify the users. Finally, we present a
brief discussion.
II. MEDICAL ASPECTS OF EEG
The electrical currents in the brain was discovered in 1875
by an English physician Richard Caton. He observed the EEG
from the exposed brains of rabbits and monkeys. In 1924 Hans
Berger, a German neurologist, used his ordinary radio
equipment to amplify the brain's electrical activity measured on
the human scalp. He announced that weak electric currents
generated in the brain can be recorded without opening the
skull, and depicted graphically on a strip of paper. The activity
that he observed changed according to the functional status of
the brain, such as in sleep, anesthesia, lack of oxygen and in
certain neural diseases, such as in epilepsy. [3]
EEG signals are generated from activities in the neurons.
When the neurons are activated, local current flows are
produced. EEG measures mostly the currents that flow during
synaptic excitations of the dendrites of many pyramidal
neurons in the cerebral cortex. Differences of electrical
potentials are caused by summed postsynaptic graded
potentials from pyramidal cells that create electrical dipoles
between soma (body of neuron) and apical dendrites (neural
branches). See figure 1.
Brain electrical current consists mostly of Na+, K+, Ca++,
and Cl- ions that are pumped through channels in neuron
membranes in the direction governed by membrane potential
[4]. The detailed microscopic picture is more sophisticated,
including different types of synapses involving variety of
neurotransmitters. Only large populations of active neurons can
generate electrical activity recordable on the head surface.
Between electrode and neuronal layers current penetrates
through skin, skull and several other layers. Weak electrical
signals detected by the scalp electrodes are massively
amplified, and then displayed on paper or stored to computer
memory [5]. The human brain electric activity starts around the
17-23 week of prenatal development. It is assumed that at birth
the full number of neural cells is already developed, roughly
1011 neurons [6]. This makes an average density of 104
neurons per cubic mm. Neurons are mutually connected into
neural nets through synapses. Adults have about 500 trillion
(5.1014) synapses. The number of synapses per one neuron
with age increases, however the number of neurons with age
decreases, thus the total number of synapses decreases with age
too. From the anatomical point of view, the brain can be
divided into three sections: cerebrum, cerebellum, and brain
stem (see figure 2). The cerebrum consists of left and right
hemisphere with highly convoluted surface layer called
cerebral cortex. The cortex is a dominant part of the central
nervous system. The cerebrum obtains centres for movement
initiation, conscious awareness of sensation, complex analysis,
and expression of emotions and behaviour. The cerebellum
coordinates voluntary movements of muscles and balance
maintaining. The brain stem controls respiration, heart
regulation, biorythms, neurohormone and hormone secretion,
etc.[5]. The highest influence to EEG comes from electric
activity of cerebral cortex due to its surface position.
Figure 1. Basic structural features of a neuron, highlighting the three principal
functions elements.
Figure 2. Regions of Human Brain, highlighting the major anatomical
divisions (the Lobes).
(http://msnowe.files.wordpress.com/2009/06/brain_witelson1.jpg).
The 8th International Conference on INFOrmatics and Systems (INFOS2012) – 14-16 May
Bio-inspired Optimization Algorithms and Their Applications Track
Faculty of Computers and Information - Cairo University BIO-57
. III. EEG signals are sinusoidal waves, their amplitude is
normally between 0.5 and 100 μV. After applying a Fourier
transform to the row signals and the power spectrum is
generated, we have four groups of waves [7] see figure 3:
Alpha 8-13 Hz, appears during relaxation without
attention and concentration.
Theta 4-7.5 Hz, theta waves in adults while awake is
abnormal, they are generated by access to unconscious
material, deep inspiration and meditation
Beta 14-26 Hz Usual working rhythm
Delta 0.5-4, usually happens during deep sleep
Figure 3.Samples of EEG demonstrating the major frequency bands that
are presented at various stages of cognitive activity.
III. EXPERIMENT ASPECTS
A. Task Definition
Each part of the brain is responsible for a certain mental or
physical activity. The identification technique will require that
the user perform a certain mental task. Accordingly this task
will trigger neurons in a certain parts of the brain that is
responsible for handling such task. Different tasks have been
used; the most used task was just letting the user stay still in a
quite environment and relax; then capture the signal for a
period of time to identify the user. Several tasks were later
introduced; R. Palaniappan used the following five different
tasks in his experiment [8]:
Baseline task. The subjects were asked to relax and
think of nothing in particular. This task was used as a
control and as a baseline measure of the EEG signals.
Geometric figure rotation task. The subjects were
given 30s to study a particular three-dimensional
block object, after which the drawing was removed
and the subjects were asked to visualize the object
being rotated about an axis. The EEG signals were
recorded during the mental rotation period.
Math task. The subjects were given nontrivial
multiplication problems, such as 79 times 56 and
were asked to solve them without vocalizing or
making any other physical movements. The tasks
were non-repeating and designed so that an
immediate answer was not apparent. The subjects
verified at the end of the task whether or not he/she
arrived at the solution and no subject completed the
task before the end of the 10s recording session.
Mental letter composing task. The subjects were
asked to mentally compose a letter to a friend or a
relative without vocalizing. Since the task was
repeated for several times the subjects were told to
continue with the letter from where they left off.
Visual counting task. The subjects were asked to
imagine a blackboard and to visualize numbers being
written on the board sequentially, with the previous
number being erased before the next number was
written. The subjects were instructed not to verbalize
the numbers but to visualize them. They were also
told to resume counting from the previous task rather
than starting over each time.
B. Signal Capturing
A typical EEG Signal capturing device consists of
electrodes with conductive media, filters and amplifiers and
analogue/digital converters. Devices have up to 256
electrodes; nowadays exists commercial devices with much
less electrodes, devices area available in with 4 or even 2
electrodes. Of course they capture much less information but
data analyzing techniques are enhanced to be able to cope with
such devices.
Electrodes are usually placed on the scalp using the 10-20
standards (see Figure 4). This method was developed to ensure
standardized reproducibility so that a subject's studies could be
compared over time and subjects could be compared to each
other. This system is based on the relationship between the
location of an electrode and the underlying area of cerebral
cortex. The "10" and "20" refer to the fact that the actual
distances between adjacent electrodes are either 10% or 20%
of the total front-back or right-left distance of the skull.
Each site has a letter to identify the lobe and a number to
identify the hemisphere location. The letters F, T, C, P and O
stand for Frontal, Temporal, Central, Parietal and Occipital
respectively. Note that there exists no central lobe, the "C"
letter is only used for identification purposes only. A "z"
(zero) refers to an electrode placed on the midline. Even
numbers (2,4,6,8) refer to electrode positions on the right
hemisphere, whereas odd numbers (1,3,5,7) refer to those on
the left hemisphere. [3]
The 8th International Conference on INFOrmatics and Systems (INFOS2012) – 14-16 May
Bio-inspired Optimization Algorithms and Their Applications Track
Faculty of Computers and Information - Cairo University BIO-58
Figure 4. Standard electrode map, illustrating the commonly deployed 10-20
System (http://www.immrama.org/eeg/electrode.html)
Values at each channel is calculated by finding the
difference in the reading of two electrodes based on one of the
below methods
Bipolar: calculating the difference related to a nearest
electrode.
Referential: calculating the difference with reference
to the ear lobe so nodes at the right side of the head
use reference of the right ear, and the left with the left
ear.
Common reference, calculating the difference to a
single electrode reference for all electrodes.
Based on the mental tasks that the user will perform,
activities should be expected at certain channels which reflect
the part of the brain responsible for this activity.
IV. PROCESSING TECHNIQUES
In this section we will discuss various techniques used in
human identification based on EEG Signals
A. Fisher’s Discriminant Analysis
A. Riera et al. [9] have developed a multimodal
authentication algorithm based on EEG and ECG signals.
They conducted the test on 40 healthy subjects. Each subject
was required to sit in a comfortable armchair, to relax, be quiet
and close their eyes. Then three 3 minute takes are recorded to
32 subjects and four 3minutes takes are recorded to the 8
subjects. The 32 subject set are used as reference subject in the
classification stage and the 8 subjects are the ones that are
enrolled into the systems. Then several 1 minute takes are
recorded afterwards to these enrolled subjects, in order to use
them as authentication tests. Two electrodes were used to
capture the EEG signals and 2 for the ECG. The data was
divided to four seconds epochs. Two types of Features were
extracted from the four seconds epochs, one channel features
(Auto regression, Fourier Transform) and Synchronicity
features. Three features were selected from the Synchronicity
features namely; Mutual information (measures the
dependency degree between two random variables given in
bits, when logarithms of base 2 are used in its computation) ,
Coherence (quantizes the correlation between two time series
at different frequencies ), Correlation measures (measure of
the similarity of two signals,). The classifier used in the
authentication process is the classical Fisher’s Discriminant
Analysis, Four different discriminant functions were used
(Linear, Diagonal Linear, quadratic, diagonal quadratic). The
five best classifiers from the original 28 classifiers generated
for each subject are selected during the enrollment and
authentication of each subject.
The False Acceptance Rate (FAR) is computed taking into
account both the intruder and the impostor cases (21.8%). The
True Acceptance Rate (TAR) only takes into account the legal
cases.(71.9%)
After combining the 2 signals (EEG and ECG) the TAR is
97.9% and the FAR is 0.82%.
B. Linear Discriminant Classifier
R Palaniappan[8] proposed a multiple mental thought
identification modal. The experiment was conducted on four
subjects The subjects were seated in an Industrial Acoustics
Company sound controlled booth with dim lighting and noise-
less fan (for ventilation). An Electro-Cap elastic electrode cap
was used to record EEG signals from positions C3, C4, P3, P4,
O1 and O2 defined by the 10-20 system of electrode
placement. Each subject was requested to do up to five mental
tasks. Signals were recorded for 10 seconds during each task
and each task was repeated 10 times. Each recording was
segmented into 20 segments, each 0.5s length. The five
mental tasks performed by the subjects are baseline task
(relaxed state), geometric figure rotation, math task (2 digit
multiplication), mental letter composing task, and visual
mental counting. The captured data features were extracted
using AR modeling. Six AR coefficients were obtained for
each channel, giving a total of 36 feature vector for each EEG
segment for a mental thought. When two mental thoughts were
used, the size of the feature vector was 72 and so forth when
more mental thoughts were used.
Linear Discriminant Classifier was used to classify the
EEG feature vectors, LDC is a linear classification method
that is computationally attractive as compared to other
classifiers like artificial neural network. Various results were
presented showing the error rate using 1,2…5 five
combination of the mental tasks. Using 1 task an average of
error rate is 2.6%, while using the 5 mental tasks, the error rate
was 0.1%.
The 8th International Conference on INFOrmatics and Systems (INFOS2012) – 14-16 May
Bio-inspired Optimization Algorithms and Their Applications Track
Faculty of Computers and Information - Cairo University BIO-59
C. LVQ Nerual Net
Cempírek et al. [10], proposed neural network
classification technique for user identification. The algorithm
was conducted on a datasets of 8 subjects. The subject was sat
is a dim and silent room, eyes kept closed. Then the EEG
recordings were segmented (segment length 180 sec, step 22.5
sec); the single segments were centered. Linear magnitude
spectra of the single segments were computed by Fast Fourier
transform (Hamming window was used).
The LVQ neural network is a self–organizing neural
network, with added second layer for vectors classification
intended to be used with unlabelled training data. The first
network layer detects subclasses. The second layer combines
these subclasses into one single class. Actually, the first layer
computes distance between input and stored patterns; the
winning neuron is the one with minimum distance. Hence
LVQ network is a kind of nearest-neighbour classifier; it does
not make clusters, but the algorithm search through the
weights of connections between input layer neurons and
output map neurons. These represent classes. The best
classification rate was around 80% .
D. Neural Network
Sun [11] has developed a user identification system based
on Neural Networks. The system was tested on 9 subjects. The
task was to imagine moving his or her left or right index finger
in response to a highly predictable visual cue. EEG signals
were recorded with 59 electrodes mounted according to the
international 10-10 system. Only Signals from 15 electrodes
were used in the system. Totally 180 trials were recorded for
each subject. Ninety trials with half labeled left and the other
half right were used for training, and the other 90 trials were
for testing. Each trial lasted six seconds with two important
cues. The preparation cue appeared at 3.75 s indicating which
hand movement should be imagined, and the execution cue
appeared at 5.0 s indicating it was time to carry out the
assigned response. The common spatial patterns (CSP) is
employed to carry out energy feature extraction. As a result,
each trial is modeled by an 8-dimensional vector (4 sources
from each kind of mental task is assumed in this paper). Based
on these features, neural network classifiers can be learned.
Neural networks of one hidden layer and one output layer for
experiments. The results showed that imagining left index
finger movements is more appropriate for personal
identification. Left index movement gave a classification
accuracy of 95.6% and right index accuracy gave 94.81%. To
summarize the above mentioned techniques’, Table I presents
a summary of these techniques.
V. CONCLUSION
This paper has summarised several techniques and
implementations that have been published with respect to the
deployment of EEG for human authentication and/or
identification. The results are quite promising – with upwards
of 95% accuracy. One of the major obstacles for deploying
the EEG as a biometric is the signal acquisition process. In
the clinical arena, large electrode caps with 128 or 256
electrodes are deployed. These caps require a significant
effort to put on and setup. The impedance must be at a
certain level, the skin requires preparation, conductive gels are
often deployed to enhance the SNR. These steps preclude the
deployment of EEG based approaches to biometrics for
obvious reasons. Furthermore, the electrode wires are
attached to cumbersome ADC boards, attached to a dedicated
computer system. What is required is a small footprint,
portable device – and these are being developed at a very rapid
pace. As indicated in table I, authentication can be quite
successful with a small number of electrodes( in fact 2 have
provided significant success in several reported cases).
Furthermore, dry electrodes are now becoming common place
– these do not require significant skin preparation and obviate
the need for messy conductive gels. Furthermore, an electrode
cap is not required – the electrodes can be placed in a
headband or a baseball cap. In addition, these devices
transmit the data wirelessly (BlueTooth of WiFi) – making
data collection a common place task. So the technology is
developing – and will continue to do so, provided there is a
market.
The studies selected for presentation in this survey are
typical – and one notices that most studies deploy a small
cohort of subjects. This is a significant design flaw – though
it is not specific to biometrics. These studies should be
reproduced on much larger scales, utilizing 100’s of subjects
within a study, in order to investigate the scalability of the
classification ability of the EEG. If these studies provide
results consistent with small scale studies, then scalability will
not be a significant factor in the deployment of EEG based
biometrics (this is a major assumption in Cognitive biometrics
[12]). Other salient issues are the temporal stability of the
EEG, and the effect of variations in mental state – such as
stress and general arousal levels. These are the principal
topics of the cognitive biometrics domain – which serves to
investigate how cognitive and emotional states, in conjunction
with basic genetic variability can provide unique signatures
that will serve as an authentication vehicle. It would be
interesting to simply put on a cap, and begin interacting with
the device, without the need to remember which login ID and
password is required for the current system. The system
would present challenges to the user, to which they would
respond naturally to – and based on the stimulus-response
protocol, the user would be authenticated. Lastly, this
approach provides a suitable mechanism for static and
continuous authentication.
The 8th International Conference on INFOrmatics and Systems (INFOS2012) – 14-16 May
Bio-inspired Optimization Algorithms and Their Applications Track
Faculty of Computers and Information - Cairo University BIO-60
TABLE I. A SUMMARY OF SELECTED STUDIES DEPLOYING EEG FOR PERSON AUTHENTICATION, INCLUDING TASK MEASURE AND THE RESULTING CLASSIFICATION
ACCURACY (TAR = TRUE ACCEPTANCE RATE AND FAR = FALSE ACCEPTANCE RATE). THE REFERENCES ARE LISTED IN THE RIGHT-MOST COLUMN.
Technique
Channels
Subjects
Task
TAR
FAR
A
2
40
Rest
79.2%
21.8%
[9]
B
6
4
Rest, Math,
Letter, Count,
Rotation
-
0.1% avg
combination
using 5
features
[8]
C
-
8
Rest
80%
[10]
D
15
9
Left/Right Hand
Movement
95.6% (left)
94.81 (Right)
[11]
REFERENCES
[1] D. J. A. Smit and D. Posthuma and D. I. Boomsma and E. J. C. De Geus,
“Heritability of background EEG across the power spectrum”
Psychophysiology Journal, vol. 42, pp. 691-697, 2005.
[2] Julie Thorpe and P.C. van Oorschot and Anil Somayaji,” Pass-thoughts:
Authenticating with Our Minds”, In Proceedings of New Security
Paradigns Workshop. Lake Arrowhead, pp.45–56, 2005.
[3] Teplan, M, “Fundamentals of EEG Measurement,” in Measurement
Science Review, vol. 2, pp. 1–11, 2002.
[4] H. L. Atwood, W. A. MacKay., “Essentials of neurophysiology,” B.C.
Decker, Hamilton, Canada, 1989.
[5] F. S. Tyner, J. R.Knott., “Fundamentals of EEG technology Volume 1:
Basic concepts andmethods,”, Raven press, New York., 1989.
[6] Y P. L. Nunez., “Neocortical Dynamics and Human EEG Rhythms,”
Oxford University Press, New York,1995.
[7] http://emedicine.medscape.com/article/1139332-overview [20100712].
[8] Ramaswamy Palaniappan,“Multiple Mental Thought Parametric
Classification: A New Approach for Individual Identification”,
Proceeding of International Journal of Signal Processing, Volume 2, PP
222-226, 2006.
[9] Alejandro Riera, Aureli Soria-Frisch, Marco Caparrini, Ivan Cester, and
Giulio Ruffini, “Multimodal Physiological Biometrics Authentication”,
Biometrics: Theory, Methods, and Applications, Wiley Press, 2010.
[10] Cempírek, M. - Šťastný, J. “The optimization of the EEG-based biometric
classification.” Applied Electronics., pp. 25-28. 2007.
[11] Shiliang Sun. “Multitask Learning for EEG-Based Biometrics”,
Proceeding of International Conference on Patter Recognition, , PP 51-55
,2008
[12] Revett, K., Cognitive Biometrics: A Novel Approach to Person
Authentication, ch 7, in Continuous Authentication using Biometrics:
Data, Methods, and Models, (eds) I. Traore and A. Ahmed, Information
Science Reference, ISBN: 1613501293, publication date: 30 September,
2011.