EEG authentication system using fuzzy vault scheme

Abstract Authentication has become a necessity these days because the use in most of the life applications from accounts to devices, especially with the rapid development of technology versus traditional methods such as password and PIN, that is, become infeasible for such methods. New methods have emerged, including biometric systems. In this paper we will explore the possibility of combining cryptography with biometrics in order to achieve authentication. EEG will be used as they are unique and also difficult to expose and copy of 9 healthy persons, extracting features from them and use that features with fuzzy vault scheme to provide high security to encrypt Biometric systems. the calsssification gives good accuracy 96% , and using the tent chaff points give the system an advantage because it reduces the error occurs when separates chaff points from the genuine point which are the EEG signal features because the initial seeds are known by both sender and receiver.


Introduction
Efficient authentication way depend on using human unique characteristics. Biometric authentications are more effective and less prostrate to such problems because of their uniqueness to each user. Biometrics refer to physical or behavioral human characteristics used to identify a person to allow access to systems, devices or data [1]. For authentication purpose, retina, iris, fingerprint, voice, and face are examples of biometrics been introduced also called conventional biometric [2]. EEG is the measurement of electrical activity of the brain, sensor used to obtain these signals. Brain consists of millions of neuron and these neuron express emotions and thoughts as signals [3].
An authentication scheme that uses cognitive and emotional state of a user to authenticate and/or identify users. These perspectives are extracted by recording EEG signals of a person. Biometrics-based user authentication has advantages over traditional authentication systems However, new authentication structure known as crypto-biometric systems, where biometric and cryptography are binding to achieve high level security [2] .
This system is so called fuzzy vault. Fuzzy vault is a scheme used as effective technique for securing biometrics, it based on using features of the biometrics and secret key which has several advantages over passwordbased systems. By Choosing secret key and construct a polynomial from it, evaluate these features on the polynomial then add random points called chaff points to produce vault this procedure called encoding, to decode vault, a set of features to be used to obtain the key are matched then reconstruct the polynomial then get the secret key, if features don't match then the polynomial can't be reconstructed. This is called decoding [4].

Related work
In 2002, Juels and Sudan [5] were the first who introduced the first concept of the fuzzy vault and they proposed the locking and unlocking algorithms of the fuzzy vault.
Uludag et al. [4] introduce fingerprint biometric combined with fuzzy vault algorithm and proposed a fingerprint-based fuzzy vault.
Kennet fladby [6] record (EEG) signals 12 participants performing eight different tasks in three sessions. then analyze signals in both the time domain and the frequency domain and extract features and propose Dynamic Time Warping as well as a feature based distance metric .
Nguyen et al [8] verify person using motor imagery eeg signals . Two tasks was performed first task: left hand or right hand in the second, only the best motor imagery task for each person was performed.and modeling usingThe Gaussian mixture model (GMM) and support vector data description (SVDD) methods. Damasevicius et al [3] present a biometric authentication method based on the (BCH) codes and discrete logarithm, and using a collected (EEG) data from 42 participant to evaluate a biometric cryptosystem ,The results show that the authentication system is effective.

Proposed Method
Fuzzy vault scheme first proposed by juels and sudan [5] ,it is an improvement of fuzzy commitment introduced by Juels and Wattenberg [7].In [5] Alice lock secret S (encryption key) using unordered set If Bob can at least find 4 correct points, lie on p, from V, he will reconstruct p, hence, decode the secret as the polynomial coefficients (1, 2, 1). Otherwise, reconstructing incorrect p, and can't access the secret S.

Dataset description
In this work BCI Competition 2008 -Graz data set A [9] is used as our EEG dataset. The data recorded from 9 subjects ,two sessions on different days, and four classes (left, right, tongue, feet). Each session includes 6 runs, each run consist of 48 trail,12 for each class, totally 288 trial / session, where the participant set on a comfortable armchair, in the beginning at (t = 0s) a fixation cross appeared on the black screen. and short warning tones at (t = 2 s), a cue in the form of an arrow appeared, pointing randomly to one of classes to the left, right, down or up, each trial last 3 seconds A short break followed. The signals were sampled with 250 Hz and (0.5 Hz -100 Hz) bandpass-filter, An additional 50 Hz notch filter was enabled to suppress line noise.with no feedback.

EEG preprocessing and features extraction
EEG signals where filtered between 8-30 HZ which are the frequencies of Alpha and Beta bands These are EEG features that corresponds to brain's normal motor output channels and two classes are chosen for authentication. Left hand and right hand for 3 channels, (C3, CZ, C4) channels are selected to reflect the interacted channel with Motor Imagery tasks.From each channel, trials where extracted and artifact trials were removed .From each trial of a channel , power spectral density (PSD) estimation using welch method is calculated to extract features from: variance, power and energy were calculated. Total 5 features for each trial selected using feature selection algorithm where: x(t) = is the trial section signal.
These features are maps into integers because galois field deals only with ineger numbers.

Coclusion and discussion
As a biometric cryptosystem, fuzzy vault can protect both the user's templet and a secret value at the same time. This is considered to be an extremely useful feature, since it introduces new applications for biometric-based user authentication [10][11][12][13][14].

Figure 1
Confussion matrix After testing the system, it gives a good accuracy of classifying which is 96% , using the tent chaff points give the system an advantage because it reduces the error occurs when separates chaff points from the genuine point which are the EEG signal features because the initial seeds are known by both sender and receiver so, the system can regenerate the chaff points again and rise them without or less effecting the genuine points and in the traditional chaff point generation it need to keep distance from the genuine point which require more calculation which this method doesn't the figure bellow illustrate the confusion matrix.
From confusion matrix also can calculate the true positive rate and the false Negative rate as shown below: