Biometric Identification in Virtual Worlds using Biometric Fusion
Techniques
Ahmed Al-Khazzar and Nick Savage
School of Engineering, University of Portsmouth, Portsmouth, U.K.
Keywords: Biometric Fusion, Biometric Recognition, Identification, Virtual Worlds, Games, Behavioural Biometric.
Abstract: The use of virtual worlds is becoming popular in many fields such as education, economy, space, and
games. With the widespread use of virtual worlds, establishing the security of these systems becomes more
important. In this paper a behavioural biometric system is implemented to identify users of a virtual
environment. This research suggests the use of a score level fusion technique to improve the identification
performance of the system. The identification is achieved by analysing user interactions within the virtual
environments and comparing these interactions with the previously recorded interactions in the database.
The results showed that using score level biometric fusion in behavioural biometric systems similar to the
one presented in this research is a promising tool to improve the performance of these systems. The use of
biometric fusion technique enhanced the performance of the implemented biometric system up to 7.5%. An
average equal error rate of up to 22.7% was achieved in this work.
1 INTRODUCTION
A virtual world is an interactive 3D virtual
environment that visually resembles complex
physical spaces, and provides an online community
through which the users can connect, shop, work,
learn, establish emotional relations, and explore
different virtual environments. Users of a virtual
world can interact with the objects of the virtual
environments through avatars. They can perform
real world activities such as watching, hearing and
touching the virtual objects through avatars.
Virtual worlds have become very popular in
many fields such as E-learning (Dharmawansa et al.,
2011); (Gonzalez-Pardo et al., 2010), economy
(Harris and Novobilski, 2008); (Kim et al., 2002);
(Peng and Xu, 2008), space (Noor, 2010); (Romann,
2007), and games (e.g. the World of Warcraft). USA
National Aeronautics and Space Administration
agency (NASA) use virtual worlds to test the design
of equipment (Cline, 2005, p. 92). In the last few
years a large number of virtual worlds have been
developed, which share a number of characteristics
(Noor, 2010):
1. Presence and real-time chat facilities in a shared
space.
2. Persistent environment in which objects continue
to exist in the absence of users and do not disappear
when users are logged out.
3. Users are represented in the virtual world by
avatars.
4. 3D graphical environments.
Cline (2005) argued potential impacts of virtual
reality environments in human life and activity. He
predicted that virtual reality will be integrated into a
human’s daily life and techniques will be developed
to influence human behaviour, interpersonal
communication and cognition. Cline (2005) also
suggested that there will be a shift from the use of
virtual reality from mainly communications to the
use of virtual reality as an extension of the real
world and a “migration to virtual space” will result
in significant changes in economics, culture and
other aspects of human life.
Therefore the future of the technology seems to
be interconnected with the future of virtual reality as
Cline (2005) predicts. With the expansion of virtual
worlds there will be a demand for security of these
newly created virtual reality environments. Similar
to all types of systems and applications, virtual
worlds require access control mechanisms to control
the access of users to the resources of these
environments. Authentication is the key component
of any access control policy in any system. While
263
Al-Khazzar A. and Savage N..
Biometric Identification in Virtual Worlds using Biometric Fusion Techniques.
DOI: 10.5220/0004054602630269
In Proceedings of the International Conference on Security and Cryptography (SECRYPT-2012), pages 263-269
ISBN: 978-989-8565-24-2
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
almost all virtual worlds implement initial
authentication through usernames and passwords,
very few (if any) virtual worlds have mechanisms to
verify the identity of the users after the initial log in.
The importance of subsequent verification results
from the possibility of intruders seizing control from
the genuine users initially logged in to the system.
The difficulty with continuously identifying users
inside virtual worlds is that it can be obtrusive and
prevent users from easily interacting with the virtual
world. However, continuous user identity
verification can be achieved unobtrusively through
analysing user interactions with the virtual
environments. Identifying users in virtual worlds
based on their interaction with these environments
not only will be useful for continuous user
recognition, but also for verifying the identity of the
users claiming to be the genuine users of the system
and possessing the genuine user password.
Knowledge based authentication mechanisms such
as passwords are currently used in virtual worlds;
however the virtual worlds are not capable of
distinguishing between genuine users and imposters
who possess the knowledge needed to gain access to
the virtual world. In addition current virtual worlds
are not capable of determining if the current user is
the continuing genuine user (who has been
authenticated to access the system at the start of the
session) or an imposter who has seized control of the
virtual world.
In this paper we propose a behavioural biometric
identification technique that utilises user interaction
with virtual worlds. The virtual worlds are strategy-
less 3D games that are implemented for the
identification purpose in order to collect the user
actions during the game play. While proposing a
more secure biometric identification system is the
main theme of this research, the study of the human
behaviour in a virtual world can have several other
applications. Examples of such applications are
differentiating humans from machines (bots) in
online games (Golle and Ducheneaut, 2005);
(Thawonmas et al., 2008); (Yampolskiy and
Govindaraju, 2007), and finding users operating
multiple accounts in an online system (Ishikawa et
al., 2010).
To the best knowledge of the authors, there is
currently no research available which implements
behaviour based user recognition inside virtual
worlds. However there are a few studies that analyse
the behaviour of users inside virtual worlds
(Dharmawansa et al., 2011); (Gavrilova and
Yampolskiy, 2010); (Gonzalez-Pardoe t al., 2010).
2 BIOMETRIC IDENTIFICATION
2.1 Introduction
Biometric identification as defined by ISO/IEC is
the process of searching against a biometric
enrolment database to find and return the biometric
reference identifier(s) matching the submitted
biometric sample of a single individual (Standing
document 2, 2007). Biometric identification systems
are usually classified into two categories:
physiological and behavioural biometric systems.
While there has been a significant surge in the
use of physiological biometric systems for user
identification and verification in recent years, they
have not been a perfect solution. There are a large
number of known attacks against these systems. A
few security attacks have been reported in (Buthan
and Hartel, 2005); (Ratha et al., 2001). Buthan and
Hartel (2005) identified three types of spoofing
attacks to biometric systems: coercive
impersonation, replay attack, and impersonation
attack. Although there are a number of
counterattacks against spoofing by using liveness
detection methods as described in (Toth, 2005), or
using a (multi-sensor) multimodal biometric system
(Schuckers, 2002), these methods add to the
complexity and cost of the biometric system and
they are not always successful.
To overcome some of these potential security
threats, a behavioural biometric system can be used.
Behavioural biometrics is a subset of biometrics
which uses measurable properties of a person’s
actions for user recognition. The behavioural
biometrics of a user are not physically accessible, in
contradiction with the other physiological biometric
methods where the user biometric is usually
physically accessible (e.g. finger prints, iris, and
face). Therefore behavioural biometrics are more
resistant against the spoofing attacks mentioned
earlier.
The proposed system of this paper is a
behavioural biometric system, utilising algorithms
used in previous systems for user identification
inside virtual environments. However the feature
extraction techniques proposed in the paper are
novel and specifically designed to extract user
interactions with the virtual worlds.
2.2 Multimodal Biometric and
Score-level Fusion
A biometric recognition system is essentially a
pattern recognition system which works by
SECRYPT2012-InternationalConferenceonSecurityandCryptography
264
acquiring biometric samples from an individual,
extracting a biometric feature set from the acquired
samples, and comparing the feature set with the
previously recorded templates in a biometric
enrolment database (Tran et al., 2011). The
biometric feature sets extracted from the user
behaviour inside a virtual world are of very different
natures. Therefore, two or more of these biometric
features can be combined to improve the efficiency
of the system.
Different levels of biometric fusion may be
defined based on the type of the available
information. Score-level fusion is the most common
fusion technique applied, due to the trade off
between information availability and fusion
complexity (Tran et al., 2011). Drosou et al. (2012)
suggested a behavioural biometric system that uses
score level fusion to combine two biometric features
for user identification based on the spatiotemporal
analysis of human activities. This paper uses a
similar approach of score level fusion of two
biometric features extracted from user interactions in
virtual worlds.
In score-level fusion the match scores which are
generated by multiple biometric comparison
modules are combined to create a new match score.
A match score is the outcome of comparing two
feature sets extracted using the same feature
extractor (Ross and Nandakumar, 2009). Match
scores are typically categorised to two classes:
similarity scores, and distance scores, which
respectively reflect the similarity or distance of the
compared biometric samples. These scores can be
rescaled arbitrarily without affecting the
performance of the biometric system, provided that
the values are scaled in a monotonic manner (Hube,
2010).
Let X be the set of similarity scores from
biometric features extracted from different feature
extractors, and let x ∈ X. The normalised similarity
score of x can be marked by′. To normalise the set
of similarity scores, the following method can be
used to map the similarity scores to interval [0, 1).
The original distribution and characteristics of the
features will be retained as the result of the scaling
process:
′ =
∑
(1)
In this paper sum-rule-based score level
(transformation based score level) fusion technique
is used to combine the new normalised similarity
scores ′ and to create a new similarity score. This
technique is generally easier than the other score-
level fusion techniques. The procedure for sum-rule-
based fusion is stated in (Horng et al., 2009): After
computing the normalised scores (x
1
, x
2
, ..., x
m
) from
a single user (from different feature extractors), the
fused score f
S
can be calculated using the following
formula:
=
+
+⋯+
(2)
The notation
represents the weight of each
normalised score
, for i = 1, 2, ..., m. In the
experiments of this research, equal weights are used.
The newly generated fused score f
S
can be used in
the comparison process to determine the identity of
the user.
3 METHODOLOGY
3.1 Data Collection
In order to have complete control on the virtual
world and the avatar actions inside virtual world, we
implemented our own virtual environment. 3D
computer games were adopted as virtual worlds for
user identification. These 3D games are considered
as interactive 3D environments with the ability to
collect user interactions with the environments for
identification purposes. Virtual worlds can have very
different environments and to investigate the user
behaviour in diverse environments with different
user avatars and movement capabilities, three
different 3D games have been considered. These
three identification environments (3D games) are
different in two main perspectives, namely, world
constraints, and character movement. Each game has
a set of different actions that can be performed using
the computer keyboard. The three implemented
games in this research are:
• A maze game (2D Movement)
• A car game (2D Movement)
• A subracer game (3D Movement)
3.2 Design of Experiments
After developing the virtual worlds, the next step is
to run experiments to collect data from users
interacting with these virtual worlds. Each user
should play the games for a specified amount of
time, called the identification time.
In tests that were performed to identify the
approximate length of time for identification inside
the developed virtual worlds of this research, 4
minutes was found to be the maximum time before
BiometricIdentificationinVirtualWorldsusingBiometricFusionTechniques
265
the users lose their concentration inside the game.
This time also provided enough data for user
identification. Therefore each user should play a
game for the period of 4 minutes for the system to
collect one set of biometric samples from the user.
The biometric sample represents a set of avatar
interactions with the environment and other data
collected from the user during a period of 4 minutes.
The experiments of this research have been
repeated twice and at different times. In each
experiment a separate group of users were asked to
play the games four times within a one month
period. There was a gap of one year between the two
experiments. The users played each game once per
week for a total of four weeks. The total sets of
samples gathered from one user for all of the games
were 12. In the first round a total of 40 users
participated in the experiment. In the second round
and a year later, a total of 50 different users
participated in the experiment. For the first round of
experiments each of the 160 sets of samples are
compared against 159 profiles giving a total of
25,440 identification tests. Similarly for the second
round each of 200 testing sets are compared against
199 profiles for a total of 39,800 identification tests.
The biometric comparison module uses a
similarity measure algorithm to classify the
extracted biometric features. The similarity measure
algorithm allows the system to compare the newly
submitted samples with the samples in the biometric
enrolment database. Various similarity measure
algorithms have been used in behavioural
recognition systems. For the sake of analysis in this
paper, the distance similarity measure is used
(Bergadano et al., 2002).
The user set of this research were all final year
male engineering students. The dominant age group
was between 20-25 years old. There were no
constraints on the time and place of the test. The
only requirement was to supply one sample of each
game per week.
3.3 Biometric Features
Biometric features are the information (in the form
of numbers or labels) extracted from biometric
samples which can be used for comparison with
other biometric samples. During the experiments,
many parameters have been collected from the users.
The parameters are: the actions of the user inside the
virtual world, the Euclidean coordinates of the game
avatar at the time of the action, and the time duration
and delay between actions. From these parameters
different features can be extracted. For the analysis
purposes of this paper, two biometric features have
been extracted, namely actions, and time biometric
features.
During the game play the user may perform
different actions, either sequentially (one by one) or
several actions at the same time (each action
corresponds pressing one or more keys). The actions
can occur in different sequences and different
frequencies. The sequence and frequency of the
actions can be used as a biometric feature to
compare biometric samples together. Each action
starts and ends at specific times, decided by the user.
Also there could be a delay between the previous
action and the next one. The time duration and delay
between actions can be used as another biometric
feature in biometric comparisons. Also, the time
biometric feature can be extracted using two
different methods. The first method is to calculate
the time between two subsequent actions. This
method is referred to as digraph method. The second
method is to calculate the time between three
subsequent actions and can be referred to as trigraph
method. An illustration of these two methods is
shown in Figure 1. Digraphs and trigraphs are used
in keystroke biometric systems (Bergadano et al.,
2002). Digraphs are defined as the latency between
two consecutively types keys. Similarly, three
consecutively typed keys are referred to as trigraphs
in these systems.
Figure 1: Time biometric feature calculation methods.
Notations t1 to t4 (digraph) and t'1 to t'3 (trigraph) are
time feature variables for actions 1 to 4.
3.3.1 Fused Biometric Scores
To compute the fused biometric scores, Equation (1)
and (2) can be used to fuse actions and time
biometric scores. Since there are two methods to
calculate the time scores, two fused scores can be
generated. These scores can be compared to find the
more efficient feature extraction method. The result
of using digraph and trigraph feature extraction
methods and fusion technique is two biometric
scores: 1- digraph fusion score, and 2- trigraph
fusion score.
SECRYPT2012-InternationalConferenceonSecurityandCryptography
266
4 RESULTS AND ANALYSIS
Table 1 illustrates the results of identification
experiments in terms of EERs. Equal error rates are
computed based on the extracted individual features
of time and action and also the score fusion of these
two biometric features. For each experiment, EERs
from five different biometric features (scores) is
reported: 1- actions, 2- digraph time, 3- trigraph
time, 4- digraph fusion, and 5- trigraph fusion. The
first three numbers represent the performance of
system in the absence of fusion techniques. The last
two numbers represent the performance of the
system when applying digraph and trigraph fusion
techniques respectively. The first round of
experiments is identified with a 2010 label, and the
second round of experiments is identified with a
2011 label. The results show EERs between 29%-
46% for individual features and 26%-36% when
applying fusion techniques.
Table 1: Average equal error rates based on individual
biometric features and fusion scores.
Game Actions
Digraph
Time
Trigraph
Time
Digraph
Fusion
Trigraph
Fusion
maze 2010 31.5 33.4 38.4 26.6 27.6
maze 2011 32.8 31.8 33.9 26.2 27.2
car 2010 36.0 34.6 40.0 33.7 33.6
car 2011 34.9 38.4 41.6 33.5 34.8
sub.2010 29.1 37.6 45.6 27.3 33.8
sub 2011 34.4 37.9 41.5 32.6 36.1
The results from Table 1 exhibits that the
“actions feature” has a better identification
performance than the “trigraph time feature”.
However this is not the case when considering
“digraph time feature”, where the EERs are
comparable; though the “action feature” still
performs better by a small margin. The similarity in
the results of the actions and “digraph time” can be
justified by the comparable discrimination power of
the time and actions behavioural features. The
slightly better results of the actions can be depicted
by the way the time feature extractor works. The
time feature value is essentially the durations of two
or three consecutive actions. When these
consecutive actions are repeated by the user in the
same session, then there are two values for the same
single feature variable. Further repeating the action
results in multiple values for this feature variable.
Since a unique value has to be assigned to each
biometric feature variable, the possible solution will
be to use the mean of these multiple values. This
mean value might not perform well in classification
tests.
It is also interesting to analyse the reason behind
the different performances of digraph and trigraph
time features. The reason behind the better
performance of the “digraph time feature” is not
instantly clear. It could be that the digraph features
possess more behavioural attributes than trigraph
features. Assuming that the user choice of the future
actions is related to the previous actions of a user,
these results could mean that in a sequence of three
consecutive actions, the choice of the third action is
less correlated to the first action and more to the
second action.
Figure 2: Performance gain in “digraph fusion”.
Figure 3: Performance gain in “trigraph fusion”.
Using score level fusion has improved the
performance across all games and in both
experiments. Both digraph and trigraph fusion
performed well in the identification tests. Figure 2
and Figure 3 show a graphical representation of the
performance gain across the games and for digraph
and trigraph fusion methods. Results show that the
maze game benefit from fusion was the most notable
with an average of 6- 7.5% increase in performance.
Maz10 Maz10 Maz11 Maz11 Car10 Car10 Car11 Car11 Sub10 Sub10 Sub11 Sub11
0
5
10
15
20
25
30
35
40
45
Games
Equal Error Rates (%)
Digraph Time-Action Average
Digraph Fusion
Maz10 Maz10 Maz11 Maz11 Car10 Car10 Car11 Car11 Sub10 Sub10 Sub11 Sub11
0
5
10
15
20
25
30
35
40
45
Games
Equal Error Rates (%)
Trigraph Time-Action Average
Trigraph Fusion
BiometricIdentificationinVirtualWorldsusingBiometricFusionTechniques
267
The car game performance gain was between 1.5-
4.5% and the subracer game performance was
between 2- 6%. These results suggest that more
constrained environments (with restricted paths),
such as the maze virtual environment, perform better
than less constrained environments, such as car
game environments.
Figure 4: Performance comparison of “digraph fusion”
between the two experiments.
4.1 Comparison of Experiments
The results of the two experiments showed that in
general the performance of biometric fusion was
consistent across both experiments. The number of
users has not affected the results. Figure 4 shows a
comparison of the fusion performances of the two
experiments. The performance reported from the
second experiment is comparable to the first
experiment across different games. The importance
of repeating experiments comes from the fact that
the discrimination power of the behavioural traits is
not easily provable. For example in the case of
fingerprint biometric systems, it is well known that
the fingerprints of humans possess a high
discriminating potential. The same cannot be said
about behavioural biometrics because of the lower
performance of these systems. However since
extensive work has been conducted on some
behavioural systems, such as keystroke based
systems, the discrimination property of these
systems are known (e.g. (Bergadano et al., 2002)). In
the case of this research, to the best knowledge of
the authors, there is no similar system available at
the time of writing and as a result repeating the
experiments is necessary to prove the discrimination
property.
5 CONCLUSIONS
This work proposed a behavioural biometric
identification system for virtual worlds, which has
the potential to identity users of virtual worlds using
a new approach, based on their interactions with the
virtual environments. In this paper three games were
implemented to test the performance of the proposed
system and biometric score level fusion technique
has been used to improve the performance of the
system. To test the performance of the system, two
experiments have been conducted. In the first
experiment 40 users, and in the second experiment, a
year later, 50 different users participated in the
experiment.
Two biometric features namely, time and action,
are extracted using two different feature extractors
(digraph and trigraph extractors). The resulting
scores from these features are normalised and then
fused using transformation based score level fusion
method. The identification tests results showed that
using this fusion technique in this particular
biometric system improves the performance of the
system significantly. The fusion technique boosted
the equal error rates to up to 7.5%. It is
recommended to use this technique to combine
biometric scores with higher performances, since it
is well known that the performance of biometric
fusion is greatly affected by its biometric feature
component with lower performance. The results also
showed that the individual digraph features
performed better than the trigraph features. This
better performance also is reflected in the fused
scores, so that the “digraph fusion” has a better
performance than the “trigraph fusion”.
The results from the two independent
experiments were similar and consistent. This is
especially vital since in behavioural recognition
systems, it is usually difficult to find whether or not
different behavioural biometric traits possess a
discrimination power to distinguish between
different users. The suggested biometric fusion
technique in this research achieved an average equal
error rate of up to 22.7%.
REFERENCES
Bergadano, F., Gunetti, D., and Picardi, C., (2002). User
authentication through keystroke dynamics. ACM
Transactions on Information and System Security , 5
(4), 367-397.
Buthan, I. and Hartel, P., (2005). The state of the art in
abuse of biometrics. Technical Report, Centre for
Maz Maz Maz Maz Car Car Car Car Sub Sub Sub Sub
0
5
10
15
20
25
30
35
40
45
Games
Equal Error Rates (%)
Digraph Fusion 10
Digraph Fusion 11
SECRYPT2012-InternationalConferenceonSecurityandCryptography
268
Telematics and Information Technology, Enschede.
Cline, M., (2005). Power, Madness, and Immortality: the
future of virtual reality. University Village Press.
Dharmawansa, A., Nakahira, K., and Fukumura, Y.,
(2011). Develop a Monitoring Tool and Extract Facial
Expression towards the Analyzing Student Behavior in
Three Dimensional Virtual Environment. Biometrics
and Kansei Engineering (ICBAKE), 2011
International Conference on (pp. 134 -139).
Takamatsu, Japan: Abrasive Engineering Society.
Drosou, A., Ioannidis, D., Moustakas, K., and Tzovaras,
D., (2012). Spatiotemporal analysis of human
activities for biometric authentication. Computer
Vision and Image Understanding , 411 - 421.
Gavrilova, M. L., and Yampolskiy, R. V., (2010).
Applying Biometric Principles to Avatar Recognition.
Proc. Int Cyberworlds (CW) Conf (pp. 179-186).
Singapore: IEEE Computer Society.
Golle, P., and Ducheneaut, N., (2005). Preventing bots
from playing online games. Computers in
Entertainment (CIE) - Theoretical and Practical
Computer Applications in Entertainment , 3 (3), 1-10.
Gonzalez-Pardo, A., Rodriguez, F. B., Pulido, E., and
Camacho, D., (2010). Using virtual worlds for
behaviour clustering-based analysis. SMVC'10 -
Proceedings of the 2010 ACM Workshop on Surreal
Media and Virtual Cloning, Co-located with ACM
Multimedia 2010, (pp. 9-14). Firenze, Italy.
Harris, B., and Novobilski, A., (2008). Real currency
economies: Using real money in virtual worlds. 2008
International Conference on Frontiers in Education:
Computer Science and Computer Engineering, FECS
2008 (pp. 241 - 246). Las Vegas: CSREA Press.
Horng, S.-J., Chen, Y.-H., Run, R.-S., Chen, R.-J., Lai, J.-
L., and Sentosal, K., (2009). An Improved Score Level
Fusion in Multimodal Biometric Systems. Parallel
and Distributed Computing, Applications and
Technologies, 2009 International Conference on (pp.
239 -246). Higashi Hiroshima : IEEE.
Hube, J. (2010). Methods for estimating biometric score
level fusion. Biometrics: Theory Applications and
Systems (BTAS), 2010 Fourth IEEE International
Conference on (pp. 1 - 6). Washington, DC, USA:
IEEE.
Ishikawa, N., Watanabe, Y., Nishimura, R., Umemoto, K.,
Okada, Y. and Murata, M., (2010). Detection of users
suspected of using multiple user accounts and
manipulating evaluations in a community site. 2010
International Conference on Natural Language
Processing and Knowledge Engineering (NLP-KE
2010), (pp. 1-8). Beijing.
Kim, B., Barua, A., and Whinston, A., (2002). Virtual
field experiments for a digital economy: a new
research methodology for exploring an information
economy. Decision Support Systems , 215 - 231.
Noor, A., (2010). Potential of virtual worlds for remote
space exploration. Advances in Engineering Software ,
666 - 673.
Peng, H., and Xu, X., (2008). The Analysis of Arbitrage
Capital Flows between Online Virtual Game and Real
Economy Based on Risk Premium. Computational
Intelligence for Modelling Control Automation, 2008
International Conference on (pp. 1141 -1146). Vienna,
Austria: IEEE.
Ratha, N. K., Connell, J. H., and Bolle, R. M., (2001).
Enhancing security and privacy in biometrics-based
authentication systems. IBM Systems Journal , 40 (3),
614-34.
Romann, J., (2007). Projective virtual reality today and in
the future: Application of virtual worlds in space,
industry, and at the construction site with robotic
know-how to improved virtual worlds. VDI Berichte ,
55-67.
Ross, A., and Nandakumar, K., (2009). Score-Level
Fusion. In S. Z. Li, and A. K. Jain, Encyclopeida of
Biometrics (pp. 611 - 621). Springer Science+Business
Media, LLC.
Schuckers, S. A., (2002). Spoofing and anti-spoofing
measures. Information security technical report , 7 (4),
56-62.
Standing document 2. (2007, 08 22). Standing document 2,
harmonized biometric vocabulary, version 8 . ISO/IEC
JTC 1/SC 37 N 2263.
Thawonmas, R., Kashifuji, Y., and Chen, K.-T., (2008).
Detection of MMORPG bots based on behavior
analysis. Proceedings of the 2008 International
Conference on Advances in Computer Entertainment
Technology, ACE 2008, (pp. 91-94). Yokohama,
Japan.
Toth, B., (2005). Biometric liveness detection.
Information Security Bulletin, (pp. 291-297).
Tran, Q., Liatsis, P., Zhu, B., and He, C., (2011). An
approach for multimodal biometric fusion under the
missing data scenario. Uncertainty Reasoning and
Knowledge Engineering (URKE), 2011 International
Conference on (pp. 185 -188). Bali, Indonesia: IEEE.
Yampolskiy, R., and Govindaraju, V., (2007). Behavioral
biometrics for recognition and verification of game
bots. 8th International Conference on Intelligent
Games and Simulation. GAME-ON 2007, (pp. 108-
114). Ghent-Zwijnaarde, Belgium.
BiometricIdentificationinVirtualWorldsusingBiometricFusionTechniques
269
