Sei sulla pagina 1di 6

Exploiting Smart Phone Accelerometer as a Personal

Identification Mechanism
WAHEED UR REHMAN*, ASADULLAH LAGHARI*, AND ZULFIQAR MEMON*
RECEIVED ON 06.01.2015 ACCEPTED ON 28.05.2015

ABSTRACT
TThe contemporary technologies for personal identity recognition schemes like PIN or password has
been adopted and used extensively by the user community. However, these schemes have certain
drawbacks; i.e. they can be broken by using bruit force dictionary or by installing key loggers on the
system. On the other side, biometric techniques are safer as compared to password identification for
security purpose. This paper presents a new biometric security mechanism using accelerometer of
smart phone. The use of accelerometer is very common for activity recognition in the researcher
community. We have used the basic concepts of graph theory for identification mechanism. The accuracy
of identification mechanism was about 80.66%.

Key Words: Biometric Signature Identification, Pattern Recognition, Smartphone Accelerometer,


Graph Theory.

1. INTRODUCTION

W
hen some services are requested from a Another method of recognition scheme is biometric
system, a personal recognition scheme is system. It is a pattern recognition system that operates on
required to confirm the identity of a person. biometric data from an individual, extracting a set of
The purpose of such system is to ensure the access of features from the data, and by comparing this feature set
legitimate user to that service. ATM, Computer systems against the templates present in the database [2].
Handwritten signature belongs to behavioral biometrics
and mobile phones are the examples of such systems.
and is widely acceptable biometric characteristics [3]. It is
One of the available traditional technologies to confirm the characteristic of an individual to write the name or
the identity is knowledge based method (PIN or password). surname by his own hand in a unique style. However
People used to set their password based on the digits or some professional forgers have ability to fool the system
the words they can easily remember, for example the names by reproducing signature of any person.

or the birthdays of their beloved. This traditional The purpose of this paper is an effort to build an
mechanism of the use of password has been criticized due identification system that overcomes the security
to its weakness in the computer society, because such limitations of passwords. In this paper we have presented
type of passwords can easily be cracked by using guess a conceptual model of security across multiple systems
or using brute force dictionary [1]. by using smart phone accelerometer. An accelerometer is

* Department of Computer Science, Institute of Business Administration, Sukkur.

Mehran University Research Journal of Engineering & Technology, Volume 34 Issue S1, August, 2015 [ISSN 0254-7821]
21
Exploiting Smart Phone Accelerometer as a Personal Identification Mechanism

an essential part of smart-phone. This sensor measures Another technique of graphical password support multiple
the linear movement in three dimensions and generates a fingers for simultaneous selection of points on the image
data element that contains the value of acceleration in all [10].
three linear dimensions [4].
Another scheme named CHC (Convex Hull Click) is
It is evident from the literature that accelerometer readings
introduced by [11]. It allows the user to prove the
provide a rich source of information about the movement knowledge about its graphical password safely in an
action of mobile phone holding by the individual [5]. insecure location because users don’t want to click directly
Using accelerometer as User Interface has also been to the image of their password. Entropy issue of pass-
proposed. If we take a smart-phone in our hand, and icon is the disadvantage.
move it as we are doing our signature in the air, during
this action the accelerometer will generate the data of Another work of keystroke dynamic has been explained
acceleration due to motion of phone that can be used as [12]. In this scheme the dynamics of keystrokes have been
a security signature. studied and it assumes that the timestamp generated by
the system related to press and release event of a key can
2. RELATED WORK be processed. Keystroke dynamic can be monitored
unobtrusively when the person is keying the information.
Many biometric schemes for the security of system have
Motion base side channel attacks can be used to detect
been proposed in the recent years. Graphical password
the key typed by any user.
techniques are nowadays common and provide better
usability than text password [6]. Modern smart phone 3. SYSTEM DESCRIPTION
using android operating system adopted a technique of
Modern mobile phones are equipped with many useful
lock mechanism different from traditional PIN codes. It is
types of input sensors including camera (often multiple),
called pattern lock and it requires a pattern by drawing
microphone, gyroscope, proximity sensor, GPS,
lines on screen by the user to unlock the smart phone.
magnetometer, tri axial accelerometer etc. These sensors
User starts the pattern by touching one of the dot and
are inexpensive, powerful and have versatile use in the
swipe its finger to join other dots to form a particular
modern research. Services using these sensors of smart-
pattern. It requires minimum 4 and maximum 9 dots to create
phone are also increasing. Especially some sensors like
the pattern. It is not allowed to repeat any node on the
gyroscope or 3 axis accelerometer have a lot of use to
way of horizontal, vertical or diagonal line. Due to this
recognize the activity of a person holding that smart phone.
limitation, the total numbers of possible patterns are
Motion capturing sensors are rapidly becoming a key
389,112 [7].
function in consumer electronic devices including smart-
phone. In our experimental setup we have used tri axial
Some researchers proposed a recognition based schemes
accelerometer on Android OS architecture. MATLAB was
which combines recall and recognition aspect by showing
used for pattern analysis.
images to the user and user have to select certain points
on the image [8-9]. Some existing graphical passwords 3.1 Accelerometer
techniques have been designed for desktop computers
originally but many of them could be used for smart-phone Accelerometer is a device used to measure the acceleration
with touch screen. of a moving object. Acceleration is the change in velocity

Mehran University Research Journal of Engineering & Technology, Volume 34 Issue S1, August, 2015 [ISSN 0254-7821]
22
Exploiting Smart Phone Accelerometer as a Personal Identification Mechanism

(speed) per unit time. Tri axial accelerometer measures the 4. SIGNATURE IDENTIFICATION
acceleration along x (lateral), y (longitudinal) and z (vertical) SYSTEM
axes (Fig. 1). It is a powerful side channel that can also be
Signature identification system can also be implemented
used even the location based services are disabled on the
for online security purpose. It may consist of an
smartphone. The electronic accelerometer uses
identification module, which is the main module of system.
piezoelectric properties of element [13]. It behaves like a
mass on a spring. When it experiences the acceleration, This main module is responsible for the complete

the mass is displaced to the position at which spring can identification logic. System has an interface for the user to
accelerate the mass at same rate as smart phone. Now this interact with. User is asked to do the signature by moving
mechanical motion is converted into electrical signals that the mobile in the air as his hand written signature.
are sensed and measured [14]. Accelerometer generates some data against the motion of
mobile phone. Smart phone transfers this data to
Accelerometer can measure static acceleration like identification module. The data base of the system
gravitational acceleration and also the acceleration of contains data template storage. The identification module
moving body in m/s2. In a smart phone (such as HTC matches the data of signature against all the templates
Desire S) it generates its value up to 6 decimal places.
present in the data base, and find the best match. Generally

3.2 Hardware Platform there are some predefined rules that must be satisfied for
identification, if best match satisfied that rules then the
We used HTC Desire S as a hardware platform for our person is identified as legitimate user.
experiment, one of the most widely used smart phone in
the market. It uses Google’s Android as operating system. In our system when user completes the activity of
signature, the accelerometer generates a signal that
3.3 Software Setup contains three component of its reading to the system.
Let F be the function of signal, which is a three dimensional
We developed an android application on HTC Desire S for
our experiment. It allows the user to start the activity of vector. The vector can be represented as:

doing his signature. The user picks the mobile phone in


F =[x, y, z] (1)
his hand and by moving it in the air in the style of his hand
written signature. The accelerometer will send new reading Where x is the lateral component of acceleration, y is the
as it senses a change of 0.3 m/s2 in the acceleration value. longitudinal component of acceleration and z is the vertical
component of acceleration

5. DATA ACQUISITION

As we have mentioned before, that our system uses tri-


axial accelerometer for the data acquisition. Accelerometer
sends packets containing all the three co-ordinates of
acceleration. Vertical component of acceleration must have
an additional value of gravitational acceleration
approximately equal to 9.8 m/s2. If mobile is picked so that
FIG. 1. AXIS OF ACCELEROMETER OF SMART-PHONE its screen faces upward then vertical component will be z,

Mehran University Research Journal of Engineering & Technology, Volume 34 Issue S1, August, 2015 [ISSN 0254-7821]
23
Exploiting Smart Phone Accelerometer as a Personal Identification Mechanism

but we observed that 9 out of 10 participants performed Where DXn ,DYn and DZn, where n=1, 2, 3, 4… N, are
the signature action by mobile device in such a way that explained in Equations (2) (3) and (4).
there vertical component was x component. So in that
case gravitational acceleration will be added in the x The total weight ‘Wt’ for N strokes will be the sum of all
component (Fig. 2). the weight of each stroke.

6. FEATURE EXTRACTION
Wt = W1 + W2 + W3+ ......+WN (6)
In order to continue the identification process firstly we
have to find the minimum and the maximum value of each
co-ordinate. For x co-ordinate we find Min(X) and Max(X)
and similarly for y and z axes. From these values we will
find the difference function for each axis. Difference
function for each co-ordinate axes can be written as:

DXn= Max(X)- Min(X) (2)

DYn= Max(Y)- Min(Y) (3)

DZn= Max(Z)- Min(Z) (4)

The second thing that we required is the Nstroke (Number


of Strokes) in a single signature. A stroke can be define as
the line start from a starting point where initial velocity
start from zero to an end point where deceleration occurs
and final velocity in that particular direction become zero.
(Deceleration is the negative acceleration or the rate of
decreasing of velocity). For further explanation of stroke
we take the help from graph theory (Fig. 3). FIG. 2. SIMPLIFIED ARCHITECTURE OF SIGNATURE
IDENTIFICATION SYSTEM
The above signature consist of three strokes, the first
stroke is from point A to point B. As we start from point A
and increase our speed then we decelerate and at point B,
the speed becomes zero in that direction. Similarly the
second stroke is from point B to point C and the third
stroke is from point C to point D.

Now we calculate the weight of each stroke of the graph.


In the above example the weight of first stroke (started
from point 1 and end at point 2) can be expressed as:

Wx  DX 12  DX 22  DX 32

W y  DY12  DY22  DY32


(5)
WY  DX  DX  DX
2
1
2
2
2
3 FIG. 3. CONTROL FLOW FOR STROKES

Mehran University Research Journal of Engineering & Technology, Volume 34 Issue S1, August, 2015 [ISSN 0254-7821]
24
Exploiting Smart Phone Accelerometer as a Personal Identification Mechanism

7. GRAPH NORM FEATURE will be over and the user will be recognized as a potential
signatory.
Graph norm is a real number that represent the whole
signature of the graph. In our case the ratio of total weight 11. SYSTEM RESULT
‘Wt’ to the Number of strokes ‘Nstroke’ can be consider
The proposed system for identification was tested on 20
as a graph norm (Gnorm).
users. For registration process each user has to provide
10 signatures. Total of 300 identification tests were
Wt
Gnorm  (7) performed by the live user. In 242 cases the system
N Stroke
correctly identified the user. In 45 cases wrong users
were identified. In 13 cases the system does not
The value of Gnorm for a normal person almost remains
suggested any user. So the identification accuracy was
same. This feature will enhance the value in the
about 80.66%.
identification process.
12. CONCLUSION AND SCOPE OF
8. USER REGISTRATION FUTURE WORK

For the registration of user, he should perform the In this paper, a new approach for security signature has
signature process at least 20 times. Each signature been proposed. This approach is based upon the
produces its own graph that will be stored in the data base accelerometer coordinates of smart phone, when the user
for one user. During the process of identification, all the performs the signature in air by holding smart phone in
stored graphs will be considered for best match. his hand. Since this approach uses graph theory for
identification, which has proven to be a good tool in the
9. IDENTIFICATION PROCESS biometric system. This system can also be used for on-
line identification applications. This method is not perfect
In the process of identification we use all the features one, because we have used only very basic concept of
extracted from the graph. These features will be compared graph theory for signature identification. The field of graph
against the set of signature templates present in the theory is wide, so in future some more feature of the graph
database to get the best match. This will be a time theory will include in our work to make it more perfect.
consuming process. To speed up the system we have to Furthermore we have not considered the Euler motion or
choose fast classifier so that only legitimate user; who the angular rotation of smart phone. So in future we will
successfully satisfy certain rules are allowed to continue also add the co-ordinates of gyroscope in our system to
the process of identification. make it more perfect.

10. IDENTIFICATION PROCEDURE ACKNOWLEDGEMENT

After the user performs his signature using smartphone Authors are thankful to the teachers of FFC (Fauji Fertilizer
the identification procedure will start. In first step the Company) School, who participated as volunteers for
graph norm and number of stokes are calculated and signature trials on our proposed system. Any opinion,
classified. If the system matches these values with finding, conclusion or recommendation expressed in the
templates in database then the identification procedure research article are those of authors only.

Mehran University Research Journal of Engineering & Technology, Volume 34 Issue S1, August, 2015 [ISSN 0254-7821]
25
Exploiting Smart Phone Accelerometer as a Personal Identification Mechanism

REFERENCES

[1] Magno, M.B., “Survey of User Authentication [8] Brostoff, S., and Sasse, M.A., “Are Passfaces More Usable
Mechanisms”, pp. 9, 1996. than Passwords? A Field Trial Investigation”, People
and Computers XIV—Usability or Else!, pp. 405-424,
[2] Jain, A.K., Ross, A., and Prabhakar, S., “An Introduction Springer, 2000.
to Biometric Recognition”, IEEE Transactions on
[9] Wiedenbeck, S., Waters, J., Sobrado, L., and Birget, J.-
Circuits and Systems for Video Technology, , Volume 14,
C., “Design and Evaluation of a Shoulder-Surfing
pp. 4-20, 2004.
Resistant Graphical Password Scheme”, Proceedings of
Working Conference on Advanced Visual Interfaces,
[3] Fotak, T., Koruga, P., and Baca, M., “Trends in Hand
2006, pp. 177-184.
Geometry Biometrics”, Croatia: Proceedings of Central
European Conference on Information and Intelligent [10] Ritter, D., Schaub, F., Walch, M., and Weber, M., “MIBA:
Systems, pp. 323-493, 2012. Multitouch Image-Based Authentication on
Smartphones”, Extended Abstracts on Human Factors
[4] Aviv, A.J., Sapp, B., Blaze, M., and Smith, J.M., in Computing Systems, pp. 787-792, 2013.
“Practicality of Accelerometer Side Channels on
[11] Sobrado, L., and Birget, J.-C., “Graphical Passwords”,
Smartphones”, Proceedings of 28th Annual Conference
The Rutgers Scholar, an Electronic Bulletin for
on Computer Security Applications, pp. 41-50, 2012.
Undergraduate Research, 2002.

[5] Tormashova, O., “Using Conventional Cell Phones to [12] Maiorana, E., Campisi, P., González-Carballo, N., and
Recognize Basic Physical Movements”, 2012. Neri, A., “Keystroke Dynamics Authentication for
Mobile Phones”, Proceedings of ACM Symposium on
[6] Andriotis, P., Tryfonas, T., Oikonomou, G., and Yildiz, Applied Computing, pp. 21-26, 2011.
C., “A Pilot Study on the Security of Pattern Screen-
[13] Scheeper, P., Gulløv, J.O., and Kofoed, L.M., “A
Lock Methods and Soft Side Channel Attacks”,
Piezoelectric Triaxial Accelerometer”, Journal of
Proceedings of 6 th ACM Conference on Security and
Micromechanics and Microengineering, Volume 6, pp.
Privacy in Wireless and Mobile Networks, pp. 1-6, 2013.
131-133, 1996.

[7] Aviv, A.J., Gibson, K., Mossop, E., Blaze, M., and Smith, [14] Bakshi, U.A., and Bakshi, A.V., “Instrumentation
J.M., “Smudge Attacks on Smartphone Touch Screens”, Engineering”, Technical Publications Pune, Volume 1,
WOOT, Volume 10, pp. 1-7, 2010. pp. 84-85, 2009.

Mehran University Research Journal of Engineering & Technology, Volume 34 Issue S1, August, 2015 [ISSN 0254-7821]
26

Potrebbero piacerti anche