See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/3094386

Three-Dimensional Password for More Secure Authentication

Article in IEEE Transactions on Instrumentation and Measurement · October 2008

DOI: 10.1109/TIM.2008.919905 · Source: IEEE Xplore

CITATIONS READS
53 9,458

2 authors:

Fawaz A Alsulaiman Abdulmotaleb El Saddik

King Saud University University of Ottawa
22 PUBLICATIONS 218 CITATIONS 824 PUBLICATIONS 15,329 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Abdulmotaleb El Saddik on 11 November 2014.

The user has requested enhancement of the downloaded file.

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 9, SEPTEMBER 2008 1929

Three-Dimensional Password for More

Secure Authentication
Fawaz A. Alsulaiman and Abdulmotaleb El Saddik, Senior Member, IEEE

Abstract—Current authentication systems suffer from many selection of passwords that are easy to remember and, at the
weaknesses. Textual passwords are commonly used; however, same time, are hard to guess.
users do not follow their requirements. Users tend to choose Klein [2] collected the passwords of nearly 15 000 accounts
meaningful words from dictionaries, which make textual pass-
words easy to break and vulnerable to dictionary or brute force that had alphanumerical passwords, and he reached the fol-
attacks. Many available graphical passwords have a password lowing observation: 25% of the passwords were guessed by
space that is less than or equal to the textual password space. using a small yet well-formed dictionary of 3 × 106 words.
Smart cards or tokens can be stolen. Many biometric authenti- Furthermore, 21% of the passwords were guessed in the first
cations have been proposed; however, users tend to resist using week and 368 passwords were guessed within the first 15 min.
biometrics because of their intrusiveness and the effect on their
privacy. Moreover, biometrics cannot be revoked. In this paper, Klein [2] stated that by looking at these results in a system with
we present and evaluate our contribution, i.e., the 3-D password. about 50 accounts, the first account can be guessed in 2 min
The 3-D password is a multifactor authentication scheme. To be and 5–15 accounts can be guessed in the first day. Klein [2]
authenticated, we present a 3-D virtual environment where the showed that even though the full textual password space for
user navigates and interacts with various objects. The sequence eight-character passwords consisting of letters and numbers is
of actions and interactions toward the objects inside the 3-D en-
vironment constructs the user’s 3-D password. The 3-D password almost 2 × 1014 possible passwords, it is easy to crack 25% of
can combine most existing authentication schemes such as textual the passwords by using only a small subset of the full password
passwords, graphical passwords, and various types of biometrics space. It is important to note that Klein’s experiment was in
into a 3-D virtual environment. The design of the 3-D virtual 1990 when the processing capabilities, memory, networking,
environment and the type of objects selected determine the 3-D and other resources were very limited compared to today’s
password key space.
technology.
Index Terms—Authentication, biometrics, graphical passwords, Many authentication systems, particularly in banking, re-
multifactor, textual passwords, 3-D passwords, 3-D virtual quire not only what the user knows but also what the
environment.
user possesses (token-based systems). However, many reports
[3]–[5] have shown that tokens are vulnerable to fraud, loss, or
I. I NTRODUCTION theft by using simple techniques.
Graphical passwords can be divided into two categories as
T HE DRAMATIC increase of computer usage has given
rise to many security concerns. One major security con-
cern is authentication, which is the process of validating who
follows: 1) recognition based and 2) recall based [1]. Vari-
ous graphical password schemes have been proposed [6]–[8],
you are to whom you claimed to be. In general, human authen- [10]–[12]. Graphical passwords are based on the idea that users
tication techniques can be classified as knowledge based (what can recall and recognize pictures better than words. However,
you know), token based (what you have), and biometrics (what some of the graphical password schemes require a long time
you are). to be performed. Moreover, most of the graphical passwords
Knowledge-based authentication can be further divided into can be easily observed or recorded while the legitimate user
two categories as follows: 1) recall based and 2) recognition is performing the graphical password; thus, it is vulnerable to
based [1]. Recall-based techniques require the user to repeat shoulder surfing attacks. Currently, most graphical passwords
or reproduce a secret that the user created before. Recognition- are still in their research phase and require more enhancements
based techniques require the user to identify and recognize the and usability studies to deploy them in the market.
secret, or part of it, that the user selected before [1]. One of Many biometric schemes have been proposed; fingerprints,
the most common recall-based authentication schemes used in palmprints, hand geometry, face recognition, voice recogni-
the computer world is textual passwords. One major drawback tion, iris recognition, and retina recognition are all different
of the textual password is its two conflicting requirements: the biometric schemes. Each biometric recognition scheme has its
advantages and disadvantages based on several factors such as
consistency, uniqueness, and acceptability. One of the main
Manuscript received November 15, 2006; revised February 6, 2008. drawbacks of applying biometrics is its intrusiveness upon
The authors are with the Multimedia Communications Research Laboratory
(MCRLab), School of Information Technology and Engineering, University of a user’s personal characteristic. Moreover, retina biometrical
Ottawa, Ottawa, ON K1N 6N5, Canada (e-mail: Fawaz@mcrlab.uottawa.ca; recognition schemes require the user to willingly subject their
abed@mcrlab.uottawa.ca). eyes to a low-intensity infrared light. In addition, most biomet-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. ric systems require a special scanning device to authenticate
Digital Object Identifier 10.1109/TIM.2008.919905 users, which is not applicable for remote and Internet users.

0018-9456/$25.00 © 2008 IEEE

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
1930 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 9, SEPTEMBER 2008

In this paper, we comprehensively analyze and discuss the user’s choices in Passfaces and in the Story scheme result in
3-D password [16]. The 3-D password is a multifactor au- a password space that is far less than the theoretical entropy.
thentication scheme. It can combine all existing authentication Therefore, it leads to an insecure authentication scheme.
schemes into a single 3-D virtual environment. This 3-D virtual The graphical password schema of Blonder [6] is considered
environment contains several objects or items with which the to be recall based since the user must remember selection loca-
user can interact. The type of interaction varies from one item tions. Moreover, PassPoint [10]–[12] is a recall-based graphical
to another. The 3-D password is constructed by observing password schema, where a background picture is presented and
the actions and interactions of the user and by observing the the user is free to select any point on the picture as the user’s
sequences of such actions. password (user’s PassPoint). Draw A Secret (DAS), which is
It is the user’s choice to select which type of authentication a recall-based graphical password schema and introduced by
techniques will be part of their 3-D password. This is achieved Jermyn et al. [13], is simply a grid in which the user creates a
through interacting only with the objects that acquire infor- drawing. The user’s drawings, which consist of strokes, are con-
mation that the user is comfortable in providing and ignoring sidered to be the user’s password. The size and the complexity
the objects that request information that the user prefers not to of the grid affect the probable password space. Larger grid sizes
provide. For example, if an item requests an iris scan and the increase the full password space. However, there are limitations
user is not comfortable in providing such information, the user in grid complexity due to human error. It becomes very hard
simply avoids interacting with that item. Moreover, giving the to recall where the drawing started and ended and where the
user the freedom of choice as to what type of authentication middle points were if we have very large grid sizes.
schemes will be part of their 3-D password and given the large One important type of authentication is based on who you are
number of objects and items in the environment, the number of or, in other words, biometrics. Biometric recognition systems
possible 3-D passwords will increase. Thus, it becomes much have been exhaustively studied as a way of authentication.
more difficult for the attacker to guess the user’s 3-D password. Fingerprints, palmprints, face recognition, voice recognition,
The remainder of this paper is organized as follows: and iris and retina recognition are all different methodolo-
Section II discusses related works. Section III introduces the gies of biometric recognition systems. However, some human
3-D password. It also discusses the guidelines of building the properties are vulnerable to change from time to time due to
3-D virtual environment and its possible applications. several reasons such as aging, scarring, face makeup, change
Section IV discusses the security analysis, including possible of hairstyle, and sickness (change of voice). Moreover, people
attacks and countermeasures. Section V presents the exper- tend to resist biometrics for different reasons. Some people
imental results. Finally, Section VI concludes and discusses think that keeping a copy of the user’s fingerprints is not
future work. acceptable and is a threat to the user’s privacy. In addition, some
users resist the idea of a low-intensity infrared light or any other
kind of light directed at their eyes, such as in retina recognition
II. R ELATED W ORKS
systems. Moreover, biometrics cannot be revoked, which leads
Many graphical password schemes have been proposed to a dilemma in case the user’s data have been forged. Unlike
[6]–[8], [10]–[12]. Blonder [6] introduced the first graphical other authentication schemes where the user can alter his/her
password schema. Blonder’s idea of graphical passwords is that textual password in case of a stolen password or replace his/her
by having a predetermined image, the user can select or touch token if it has been stolen or forged, a user’s biometrics cannot
regions of the image causing the sequence and the location of be revoked.
the touches to construct the user’s graphical password. After Many authentication systems are based on tangible objects
Blonder [6], the notion of graphical passwords was developed. and are referred to as token-based systems. Many token-based
Many graphical password schemes have been proposed. Exist- systems are vulnerable to theft and loss; therefore, most token-
ing graphical passwords can be categorized into two categories based systems require a personal identification number (PIN)
as follows: 1) recall based and 2) recognition based [1]. for authentication.
Dhamija and Perrig [7] proposed Déjà Vu, which is a The 3-D password [16] has been proposed, and initial results
recognition-based graphical password system that authenticates have been presented.
users by choosing portfolios among decoy portfolios. These
portfolios are art randomized portfolios. Each image is derived
III. 3-D P ASSWORD S CHEME
from an 8-B seed. Therefore, an authentication server does
not need to store the whole image; it simply needs to store In this section, we present a multifactor authentication
the 8-B seed. Another recognition-based graphical password scheme that combines the benefits of various authentication
is Passfaces [8]. Passfaces simply works by having the user schemes. We attempted to satisfy the following requirements.
select a subgroup of k faces from a group of n faces. For 1) The new scheme should not be either recall based or
authentication, the system shows m faces and one of the faces recognition based only. Instead, the scheme should be
belongs to the subgroup k. The user has to do the selection a combination of recall-, recognition-, biometrics-, and
many times to complete the authentication process. Another token-based authentication schemes.
scheme is the Story scheme [9], which requires the selection of 2) Users ought to have the freedom to select whether the 3-D
pictures of objects (people, cars, foods, airplanes, sightseeing, password will be solely recall-, biometrics-, recognition-,
etc.) to form a story line. Davis et al. [9] concluded that the or token-based, or a combination of two schemes or more.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
ALSULAIMAN AND EL SADDIK: THREE-DIMENSIONAL PASSWORD FOR MORE SECURE AUTHENTICATION 1931

This freedom of selection is necessary because users The action toward an object (assume a fingerprint recognition
are different and they have different requirements. Some device) that exists in location (x1 , y1 , z1 ) is different from the
users do not like to carry cards. Some users do not like actions toward a similar object (another fingerprint recognition
to provide biometrical data, and some users have poor device) that exists in location (x2 , y2 , z2 ), where x1
= x2 ,
memories. Therefore, to ensure high user acceptability, y1
= y2 , and z1
= z2 . Therefore, to perform the legitimate
the user’s freedom of selection is important. 3-D password, the user must follow the same scenario per-
3) The new scheme should provide secrets that are easy to formed by the legitimate user. This means interacting with the
remember and very difficult for intruders to guess. same objects that reside at the exact locations and perform the
4) The new scheme should provide secrets that are not easy exact actions in the proper sequence.
to write down on paper. Moreover, the scheme secrets
should be difficult to share with others.
B. 3-D Password Selection and Inputs
5) The new scheme should provide secrets that can be easily
revoked or changed. Let us consider a 3-D virtual environment space of size G ×
Based on the aforementioned requirements, we propose our G × G. The 3-D environment space is represented by the co-
contribution, i.e., the 3-D password authentication scheme. ordinates (x, y, z) ∈ [1, . . . , G] × [1, . . . , G] × [1, . . . , G]. The
objects are distributed in the 3-D virtual environment with
unique (x, y, z) coordinates. We assume that the user can
A. 3-D Password Overview
navigate into the 3-D virtual environment and interact with
The 3-D password is a multifactor authentication scheme. the objects using any input device such as a mouse, key-
The 3-D password presents a 3-D virtual environment con- board, fingerprint scanner, iris scanner, stylus, card reader, and
taining various virtual objects. The user navigates through this microphone. We consider the sequence of those actions and
environment and interacts with the objects. The 3-D password interactions using the previous input devices as the user’s 3-D
is simply the combination and the sequence of user interactions password. For example, consider a user who navigates through
that occur in the 3-D virtual environment. The 3-D password the 3-D virtual environment that consists of an office and a
can combine recognition-, recall-, token-, and biometrics-based meeting room. Let us assume that the user is in the virtual office
systems into one authentication scheme. This can be done and the user turns around to the door located in (10, 24, 91) and
by designing a 3-D virtual environment that contains objects opens it. Then, the user closes the door. The user then finds a
that request information to be recalled, information to be computer to the left, which exists in the position (4, 34, 18),
recognized, tokens to be presented, and biometrical data to be and the user types “FALCON.” Then, the user walks to the
verified. For example, the user can enter the virtual environment meeting room and picks up a pen located at (10, 24, 80) and
and type something on a computer that exists in (x1 , y1 , z1 ) draws only one dot in a paper located in (1, 18, 30), which is the
position, then enter a room that has a fingerprint recognition dot (x, y) coordinate relative to the paper space is (330, 130).
device that exists in a position (x2 , y2 , z2 ) and provide his/her The user then presses the login button. The initial representation
fingerprint. Then, the user can go to the virtual garage, open of user actions in the 3-D virtual environment can be recorded
the car door, and turn on the radio to a specific channel. The as follows:
combination and the sequence of the previous actions toward
the specific objects construct the user’s 3-D password. (10, 24, 91) Action = Open the office door;
Virtual objects can be any object that we encounter in real (10, 24, 91) Action = Close the office door;
life. Any obvious actions and interactions toward the real-life (4, 34, 18) Action = Typing, “F”;
objects can be done in the virtual 3-D environment toward the (4, 34, 18) Action = Typing, “A”;
virtual objects. Moreover, any user input (such as speaking (4, 34, 18) Action = Typing, “L”;
in a specific location) in the virtual 3-D environment can be (4, 34, 18) Action = Typing, “C”;
considered as a part of the 3-D password. We can have the (4, 34, 18) Action = Typing, “O”;
following objects: (4, 34, 18) Action = Typing, “N”;
1) a computer with which the user can type; (10, 24, 80) Action = Pick up the pen;
2) a fingerprint reader that requires the user’s fingerprint; (1, 18, 80) Action = Drawing, point = (330, 130).
3) a biometrical recognition device;
4) a paper or a white board that a user can write, sign, or This representation is only an example. The extensive real
draw on; representation will not be discussed in this paper. In order for
5) an automated teller machine (ATM) that requests a token; a legitimate user to be authenticated, the user has to follow the
6) a light that can be switched on/off; same sequence and type of actions and interactions toward the
7) a television or radio where channels can be selected; objects for the user’s original 3-D password. Fig. 1 shows a
8) a staple that can be punched; virtual computer that accepts textual passwords as a part of a
9) a car that can be driven; user’s 3-D password.
10) a book that can be moved from one place to another; Three-dimensional virtual environments can be designed to
11) any graphical password scheme; include any virtual objects. Therefore, the first building block
12) any real-life object; of the 3-D password system is to design the 3-D virtual
13) any upcoming authentication scheme. environment and to determine what objects the environment

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
1932 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 9, SEPTEMBER 2008

real life. Objects used in virtual environments should be

relatively similar in size to real objects (sized to scale).
Possible actions and interactions toward virtual objects
should reflect real-life situations. Object responses should
be realistic. The target should have a 3-D virtual environ-
ment that users can interact with, by using common sense.
2) Object uniqueness and distinction: Every virtual object
or item in the 3-D virtual environment is different from
any other virtual object. The uniqueness comes from
the fact that every virtual object has its own attributes
such as position. Thus, the prospective interaction with
object 1 is not equal to the interaction with object 2. How-
ever, having similar objects such as 20 computers in one
place might confuse the user. Therefore, the design of the
3-D virtual environment should consider that every object
should be distinguishable from other objects. A simple
real-life example is home numbering. Assume that there
Fig. 1. Snapshot of a proof-of-concept 3-D virtual environment, where the
user is typing a textual password on a virtual computer as a part of the user’s are 20 or more homes that look like each other and the
3-D password. homes are not numbered. It would be difficult to distin-
guish which house was visited a month ago. Similarly,
in designing a 3-D virtual environment, it should be easy
for users to navigate through and to distinguish between
objects. The distinguishing factor increases the user’s
recognition of objects. Therefore, it improves the system
usability.
3) Three-dimensional virtual environment size: A 3-D vir-
tual environment can depict a city or even the world. On
the other hand, it can depict a space as focused as a single
room or office. The size of a 3-D environment should
be carefully studied. A large 3-D virtual environment
will increase the time required by the user to perform a
3-D password. Moreover, a large 3-D virtual environment
can contain a large number of virtual objects. Therefore,
the probable 3-D password space broadens. However, a
small 3-D virtual environment usually contains only a few
objects, and thus, performing a 3-D password will take
Fig. 2. Snapshot of a proof-of-concept virtual art gallery, which contains 36 less time.
pictures and six computers. 4) Number of objects (items) and their types: Part of design-
ing a 3-D virtual environment is determining the types
will contain. In addition, specifying the object’s properties of objects and how many objects should be placed in the
is part of the system design. The design of the 3-D virtual environment. The types of objects reflect what kind of
environment influences the overall password space, usability, responses the object will have. For simplicity, we can
and performance of the 3-D password system. Fig. 2 shows a consider requesting a textual password or a fingerprint
snapshot of an experimental 3-D virtual environment. as an object response type. Selecting the right object
To simplify the idea of how a 3-D password works, Fig. 3 response types and the number of objects affects the
shows a state diagram of a possible 3-D password authentica- probable password space of a 3-D password.
tion system. 5) System importance: The 3-D virtual environment
should consider what systems will be protected by
C. 3-D Virtual Environment Design Guidelines a 3-D password. The number of objects and the types
of objects that have been used in the 3-D virtual envi-
Designing a well-studied 3-D virtual environment affects the ronment should reflect the importance of the protected
usability, effectiveness, and acceptability of a 3-D password system.
system. Therefore, the first step in building a 3-D password
system is to design a 3-D environment that reflects the adminis-
tration needs and the security requirements. The design of 3-D D. 3-D Password Applications
virtual environments should follow these guidelines. Because a 3-D password can have a password space that is
1) Real-life similarity: The prospective 3-D virtual environ- very large compared to other authentication schemes, the 3-D
ment should reflect what people are used to seeing in password’s main application domains are protecting critical

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
ALSULAIMAN AND EL SADDIK: THREE-DIMENSIONAL PASSWORD FOR MORE SECURE AUTHENTICATION 1933

Fig. 3. State diagram of a possible 3-D password application.

systems and resources. Possible critical applications include the to fit any system’s needs. A small 3-D virtual environment can
following. be used in many systems, including the following:
1) Critical servers: Many large organizations have critical 1) ATMs;
servers that are usually protected by a textual password. 2) personal digital assistants;
A 3-D password authentication proposes a sound re- 3) desktop computers and laptop logins;
placement for a textual password. Moreover, entrances 4) web authentication.
to such locations are usually protected by access cards
and sometimes PIN numbers. Therefore, a 3-D password
can be used to protect the entrance to such locations and
IV. S ECURITY A NALYSIS
protect the usage of such servers.
2) Nuclear and military facilities: Such facilities should To analyze and study how secure a system is, we have to
be protected by the most powerful authentication sys- consider how hard it is for the attacker to break such a system.
tems. The 3-D password has a very large probable pass- A possible measurement is based on the information content of
word space, and since it can contain token-, biometrics-, a password space, which is defined in [13] as “the entropy of
recognition-, and knowledge-based authentications in a the probability distribution over that space given by the relative
single authentication system, it is a sound choice for high- frequencies of the passwords that users actually choose.” We
level security locations. have seen that textual password space may be relatively large;
3) Airplanes and jetfighters: Because of the possible threat however, an attacker might only need a small subset of the full
of misusing airplanes and jetfighters for religio-political password space as Klein [2] observed to successfully break
agendas, usage of such airplanes should be protected by such an authentication system. As a result, it is important to
a powerful authentication system. The 3-D password is have a scheme that has a very large possible password space
recommended for these systems. as one factor for increasing the work required by the attacker
In addition, 3-D passwords can be used in less critical to break the authentication system. Another factor is to find
systems because the 3-D virtual environment can be designed a scheme that has no previous or existing knowledge of the

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
1934 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 9, SEPTEMBER 2008

Fig. 4. Password space of the 3-D password, textual password, Passfaces, and DAS with grid sizes of 5 × 5 and 10 × 10. Length is the number of actions and
interactions for a 3-D password, the number of characters for textual passwords, the number of selections for Passfaces, and the number of points that represent the
strokes for DAS. The length is up to eight (characters/actions, interactions, inputs/selections). The 3-D password virtual environment is as specified in Section V-A;
bit size is the log2 of the entire probable password space.

most probable user password selection, which can also resist where xi = xj , yi = yj , and zi = zj , only if i = j. The design
the attack on such an authentication scheme. of the 3-D environment will determine the value of Omax .
In Section IV-A, we will discuss the size of the 3-D password The variable m represents all possible actions and interactions
space. Then, we will study the knowledge distribution of the toward all existing objects Oi . However, g(AC) counts the total
3-D password. Afterward, we will analyze the possible attacks number of actions and inputs toward the 3-D virtual environ-
on the 3-D password. ment, whereas m, as we mentioned before, counts the actions
and interactions toward the objects. An example of g(AC) can
be a user movement pattern, which can be considered as a part
A. 3-D Password Space Size of the user’s 3-D password.
The function
One important factor to determine how difficult it is to
launch an attack on an authentication system is the size of the
password space. To determine the 3-D password space, we have h(Oi , Ti , xi , yi , zi ) = f (Oi , Ti , xi , yi , zi ) (3)
to count all possible 3-D passwords that have a certain number
of actions, interactions, and inputs toward all objects that exist is the number of possible actions and interactions toward the
in the 3-D virtual environment. We assume that the length of the object Oi based on the object type Ti . Object types can be
3-D password is Lmax , and the probability of the 3-D password textual password objects, DAS objects, or any authentication
of size greater than Lmax is zero. scheme.
The function f is determined from the object type. It counts

To measure the 3-D password space, we will calculate
(Lmax , G) on a 3-D virtual environment that has the space the possible actions and interactions that the object can ac-
(G × G × G) for a 3-D password of a length (number of cept. If we assume that an object “Keyboard” is in location
actions, interactions, and inputs) of Lmax or less. (x0 , y0 , z0 ) of type = textual password, f will count the possi-
In the following expression, AC represents the
possible ac- ble characters and numbers that can be typed, which is around
tions toward the 3-D virtual environment, whereas represents 93 possibilities. As we mentioned before, an object type is one
the total number of possible 3-D passwords of length Lmax of the important factors that affects the overall password space.
or less: Therefore, higher outcomes of function f means larger 3-D
password space size.
 
n=Lmax From the previous equations, we observe that the number
(Lmax , G) = (m + g(AC))n . (1) of objects and the type of actions and interactions determines
n=1 the probable password space. Therefore, the design of the 3-D
virtual environment is a very critical part of the 3-D password
In the following expression (2), Omax is the number of system. Figs. 4 and 5 illustrate the resulting password space
objects in the 3-D virtual environment: of the proposed 3-D password compared to textual password,
Passfaces, and DAS of a grid of 5 × 5 and 10 × 10, respectively.

i=O max Notice the difference between a 3-D password built on a simple
m= h(Oi , Ti , xi , yi , zi ) (2) 3-D virtual environment compared to the other authentication
i=1 schemes.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
ALSULAIMAN AND EL SADDIK: THREE-DIMENSIONAL PASSWORD FOR MORE SECURE AUTHENTICATION 1935

Fig. 5. Password space of the 3-D password, textual password, Passfaces, and DAS with grid sizes of 5 × 5 and 10 × 10. Length is the number of actions and
interactions for a 3-D password, the number of characters for textual passwords, the number of selections for Passfaces, and the number of points that represent the
strokes for DAS. The length is up to eight (characters/actions, interactions, inputs/selections). The 3-D password virtual environment is as specified in Section V-A;
bit size is the log2 of the entire probable password space.

Fig. 6. Observing the number of possible actions/interactions of a 3-D password within a 3-D environment specified in Section V-A compared to the two critical
points of textual passwords. Point “a” is the bit size of Klein [2] (3 × 106 ) dictionary of eight-character textual passwords. Point “b” represents the full password
space of eight-character textual passwords.

Fig. 6 shows the points where the 3-D password exceeds behind dictionary attacks. Klein [2] used such knowledge to
two important textual password points. Point “a” shows that by collect a small set of 3 × 106 words that have a high prob-
having only two actions and interactions as a 3-D password, ability of usage among users. The question is how has such
the 3-D password exceeds the number of textual passwords information (highly probable passwords) been found and why.
used by Klein [2] to break 25% of textual passwords of eight Users tend to choose words that have meaning, such as places,
characters. Point “b” represents the full textual password space names, famous people’s names, sports terms, and biological
of eight characters or less. It shows that by performing only terminologies. Therefore, finding these different words from the
four interactions, actions, and inputs as a 3-D password, the dictionary is a relatively simple task. Using such knowledge
3-D password space exceeds the full textual passwords of eight yields a high success rate for breaking textual passwords. Any
characters or less. authentication scheme is affected by the knowledge distribution
of the user’s secrets. According to Davis et al. [9], Passfaces
[8] users tend to choose faces that reflect their own taste on
B. 3-D Password Distribution Knowledge
facial attractiveness, race, and gender. Moreover, 10% of male
Studying the user’s behavior of password selection and passwords have been guessed in only two guesses. Another
knowing the most probable textual passwords are the key study [14] about user selection of DAS [13] concluded that

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
1936 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 9, SEPTEMBER 2008

for their secret passwords, users tend to draw things that have TABLE I
RESULTING NUMBER OF POSSIBLE 3-D PASSWORDS OF TOTAL LENGTH
meaning, which simplifies the attacker’s task. Lmax IN A 3-D VIRTUAL ENVIRONMENT AS SPECIFIED IN SECTION V-A
Currently, knowledge about user behaviors on selecting their
3-D password does not exist. Every user has different require-
ments and preferences when selecting the appropriate 3-D pass-
word. This fact will increase the effort required to find a pattern
of user’s highly selected 3-D password. In addition, since the
3-D password combines several authentication schemes into
a single authentication environment, the attacker has to study
every single authentication scheme and has to discover what
the most probable selected secrets are. For textual password, the
highly probable selected textual password might be determined
by the use of dictionaries. However, there are many authentica-
tion schemes with undiscovered probable password space.
Since every 3-D password system can be designed accord-
ing to the protected system requirements, the attacker has to
separately study every 3-D password system. This is because
objects that exist in one 3-D password system might not exist on
other 3-D password systems. Therefore, more effort is required
to build the knowledge of most probable 3-D passwords.

C. Attacks and Countermeasures

To realize and understand how far an authentication scheme
is secure, we have to consider all possible attack methods. We
have to study whether the authentication scheme proposed is
immune against such attacks or not. Moreover, if the proposed
authentication scheme is not immune, we then have to find the forging all existing biometrical data and may require forging
countermeasures that prevent such attacks. In this section, we token-based data. In addition, it requires a study of the user’s
try to cover most possible attacks and whether the attack is valid selection of objects, or a combination of objects, that the user
or not. Moreover, we try to propose countermeasures for such will use as a 3-D password. Moreover, a well-studied attack
attacks. is very hard to accomplish since the attacker has to perform a
Brute Force Attack: The attacker has to try all possible 3-D customized attack for every different 3-D virtual environment
passwords. This kind of attack is very difficult for the following design. Every system can be protected by a 3-D password that
reasons. is based on a unique 3-D virtual environment. This environment
1) Time required to login: The total time needed for a has a number of objects and types of object responses that differ
legitimate user to login may vary from 20 s to 2 min from any other 3-D virtual environment. Therefore, a carefully
or more, depending on the number of interactions and customized study is required to initialize an effective attack.
actions, the size of the 3-D virtual environment, and the Shoulder Surfing Attack: An attacker uses a camera to record
type of actions and interactions done by the user as a the user’s 3-D password or tries to watch the legitimate user
3-D password. Therefore, a brute force attack on a 3-D while the 3-D password is being performed. This attack is the
password is very difficult and time consuming. most successful type of attack against 3-D passwords and some
2) Cost of attacks: In a 3-D virtual environment that contains other graphical passwords. However, the user’s 3-D password
biometric recognition objects and token-based objects, may contain biometrical data or textual passwords that cannot
the attacker has to forge all possible biometric informa- be seen from behind. The attacker may be required to take
tion and forge all the required tokens. The cost of forging additional measures to break the legitimate user’s 3-D pass-
such information is very high; therefore, cracking the word. Therefore, we assume that the 3-D password should be
3-D password is more challenging. Moreover, the high performed in a secure place where a shoulder surfing attack
number of possible 3-D password spaces (as shown in cannot be performed.
Table I) leaves the attacker with almost no chance of Timing Attack: In this attack, the attacker observes how long
breaking the 3-D password. it takes the legitimate user to perform a correct sign-in using the
Well-Studied Attack: The attacker tries to find the highest 3-D password. This observation gives the attacker an indication
probable distribution of 3-D passwords. However, to launch of the legitimate user’s 3-D password length. However, this
such an attack, the attacker has to acquire knowledge of the kind of attack alone cannot be very successful since it gives the
most probable 3-D password distributions. Acquiring such attacker mere hints. Therefore, it would probably be launched
knowledge is very difficult because the attacker has to study as part of a well-studied or brute force attack. Timing attacks
all the existing authentication schemes that are used in the 3-D can be very effective if the 3-D virtual environment is poorly
environment. Moreover, acquiring such knowledge may require designed.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
ALSULAIMAN AND EL SADDIK: THREE-DIMENSIONAL PASSWORD FOR MORE SECURE AUTHENTICATION 1937

V. E XPERIMENTAL R ESULTS properties, and some other authentication schemes are based
on user’s knowledge such as textual and graphical pass-
We have built an experimental 3-D virtual environment that
words. Moreover, there are some other important authentication
contains several objects of two types. The first type of response
schemes that are based on what you have, such as smart cards.
is the textual password. The second type of response is re-
Among the various authentication schemes, textual password
questing graphical passwords. Almost 30 users volunteered to
and token-based schemes, or the combination of both, are com-
experiment with the environment. We asked the users to create
monly applied. However, as mentioned before, both authentica-
their 3-D password and to sign-in using their 3-D password
tion schemes are vulnerable to certain attacks. Moreover, there
several times over several days.
are many authentication schemes that are currently under study
and they may require additional time and effort to be applicable
A. Experimental Virtual 3-D Environment for commercial use.
The 3-D password is a multifactor authentication scheme that
In our experiment, we have used Java Open GL to build
combines these various authentication schemes into a single
the 3-D virtual environment and we have used a 1.80-GHz
3-D virtual environment. The virtual environment can contain
Pentium M Centrino machine with 512-MB random access
any existing authentication scheme or even any upcoming
memory and ATI Mobility Radeon 9600 video card.
authentication schemes by adding it as a response to actions
The design of the experimental 3-D virtual environment
performed on an object. Therefore, the resulted password space
represents an art gallery that the user can walk through and is
becomes very large compared to any existing authentication
depicted in Fig. 2.
schemes.
The design of the 3-D virtual environment, the selections
B. User Study of objects inside the environment, and the object’s type reflect
the resulted password space. It is the task of the system ad-
We conducted a user study on 3-D passwords using the ministrator to design the environment and to select the appro-
experimental 3-D virtual environments. The study reviewed the priate object that reflects the protected system requirements.
usage of textual passwords and other authentication schemes. Additionally, designing a simple and easy to use 3-D virtual
The study covered almost 30 users. The users varied in age, sex, environment is a factor that leads to a higher user acceptability
and education level. Even though it is a small set of users, the of a 3-D password system.
study produced some distinct results [13], [15]. We observed The choice of what authentication schemes will be part of
the following regarding textual passwords, 3-D passwords, and the user’s 3-D password reflects the user’s preferences and
other authentication schemes. requirements. A user who prefers to remember and recall a
1) Most users who use textual passwords of 9–12 character password might choose textual and graphical passwords as
lengths or who use random characters as a password have part of their 3-D password. On the other hand, users who
only one to three unique passwords. have more difficulty with memory or recall might prefer to
2) More than 50% of user’s textual passwords are eight choose smart cards or biometrics as part of their 3-D password.
characters or less. Moreover, users who prefer to keep any kind of biometrical data
3) Almost 25% of users use meaningful words as their private might not interact with objects that require biometric
textual passwords. information. Therefore, it is the user’s choice and decision to
4) Almost 75% of users use meaningful words or partially construct the desired and preferred 3-D password.
meaningful words as their textual passwords. In contrast, The 3-D password is still in its early stages. Designing
only 25% of users use random characters and letters as various kinds of 3-D virtual environments, deciding on pass-
textual passwords. word spaces, and interpreting user feedback and experiences
5) Over 40% of users have only one to three unique textual from such environments will result in enhancing and improving
passwords, and over 90% of users have eight unique the user experience of the 3-D password. Moreover, gathering
textual passwords or less. attackers from different backgrounds to break the system is one
6) Over 90% of users do not change their textual passwords of the future works that will lead to system improvement and
unless they are required to by the system. prove the complexity of breaking a 3-D password. Moreover, it
7) Over 95% of users under study have never used any will demonstrate how the attackers will acquire the knowledge
graphical password scheme as a means of authentication. of the most probable 3-D passwords to launch their attacks.
8) Most users feel that 3-D passwords have a high Shoulder surfing attacks are still possible and effective
acceptability. against 3-D passwords. Therefore, a proper solution is a field
9) Most users believe that there is no threat to personal of research.
privacy by using a 3-D password as an authentication
scheme.
R EFERENCES
[1] X. Suo, Y. Zhu, and G. S. Owen, “Graphical passwords: A
VI. C ONCLUSION AND F UTURE W ORK survey,” in Proc. 21st Annu. Comput. Security Appl. Conf., Dec. 5–9,
2005, pp. 463–472.
There are many authentication schemes in the current state. [2] D. V. Klein, “Foiling the cracker: A survey of, and improvement to pass-
Some of them are based on user’s physical and behavioral words security,” in Proc. USENIX Security Workshop, 1990, pp. 5–14.

Authorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.
1938 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 9, SEPTEMBER 2008

[3] NBC news, ATM Fraud: Banking on Your Money, Dateline Hidden Abdulmotaleb El Saddik (M’02–SM’03) received
Cameras Show Criminals Owning ATMs, Dec. 11, 2003. the Dipl.Ing. and Dr.Ing. degrees from the Technical
[4] T. Kitten, Keeping an Eye on the ATM. (2005, Jul. 11). [Online]. University of Darmstadt, Darmstadt, Germany, in
Available: ATMMarketPlace.com 1995 and 2001, respectively.
[5] BBC news, Cash Machine Fraud up, Say Banks, Nov. 4, 2006. He is the University Research Chair and an As-
[6] G. E. Blonder, “Graphical password,” U.S. Patent 5 559 961, sociate Professor with the School of Information
Sep. 24, 1996. Technology and Engineering, University of Ottawa,
[7] R. Dhamija and A. Perrig, “Déjà Vu: A user study using images for Ottawa, ON, Canada, the Director of the Multimedia
authentication,” in Proc. 9th USINEX Security Symp., Denver, CO, Communications Research Laboratory (MCRLab),
Aug. 2000, pp. 45–58. and the Director of the Information and Commu-
[8] Real User Corporation, The Science Behind Passfaces. (2005, Oct.). nications Technology Cluster, Ontario Network on
[Online]. Available: http://www.realusers.com E-Commerce (ORNEC). He is a Theme Coleader in the Learning Objects
[9] D. Davis, F. Monrose, and M. K. Reiter, “On user choice in graphical Repository Network (LORNET) NSERC Research Network. He is an Associate
password schemes,” in Proc. 13th USENIX Security Symp., San Diego, Editor of the ACM Transactions on Multimedia Computing, Communications
CA, Aug. 2004, pp. 1–14. and Applications. He has authored or coauthored two books and more than
[10] S. Wiedenbeck, J. Waters, J.-C. Birget, A. Brodskiy, and N. Memon, 180 publications. He is a leading researcher in haptics, service-oriented ar-
“Authentication using graphical passwords: Effects of tolerance and im- chitectures, collaborative environments, and ambient interactive media and
age choice,” in Proc. Symp. Usable Privacy Security, Pittsburgh, PA, communications.
Jul. 2005, pp. 1–12. Prof. El Saddik is an IEEE Distinguished Lecturer and has served as a Guest
[11] S. Wiedenbeck, J. Waters, J.-C. Birget, A. Brodskiy, and N. Memon, “ Editor for several IEEE TRANSACTIONS and journals. He was a recipient of
Authentication using graphical passwords: Basic results,” in Proc. the Premier’s Research Excellence Award and the National Capital Institute of
Human-Comput. Interaction Int., Las Vegas, NV, Jul. 25–27, 2005. Telecommunications New Professorship Incentive Award in 2004, the Friedrich
[12] S. Wiedenbeck, J. Waters, J.-C. Birget, A. Brodskiy, and N. Memon, Wilhelm-Bessel Research Award from the Alexander von Humboldt Founda-
“PassPoints: Design and longitudinal evaluation of a graphical password tion, Germany, in 2007, and the Best Paper Awards from the Sixth International
system,” Int. J. Human-Comput. Stud. (Special Issue on HCI Research in Conference on Virtual Concepts and the Third IEEE International Workshop
Privacy and Security), vol. 63, no. 1/2, pp. 102–127, Jul. 2005. on Collaborative Service-Oriented P2P Information Systems in 2006 and 2007,
[13] I. Jermyn, A. Mayer, F. Monrose, M. K. Reiter, and A. D. Rubin, “The de- respectively.
sign and analysis of graphical passwords,” in Proc. 8th USENIX Security
Symp., Washington DC, Aug. 1999, pp. 1–14.
[14] J. Thorpe and P. C. van Oorschot, “Graphical dictionaries and the
memorable space of graphical passwords,” in Proc. USENIX Security,
San Diego, CA, Aug. 9–13, 2004, p. 10.
[15] A. Adams and M. A. Sasse, “Users are not the enemy: Why users compro-
mise computer security mechanisms and how to take remedial measures,”
Commun. ACM, vol. 42, no. 12, pp. 40–46, Dec. 1999.
[16] F. A. Alsulaiman and A. El Saddik, “A novel 3D graphical password
schema,” in Proc. IEEE Int. Conf. Virtual Environ., Human-Comput.
Interfaces, Meas. Syst., Jul. 2006, pp. 125–128.

Fawaz A. Alsulaiman received the B.S. degree

in computer sciences from King Saud University
(KSU), Riyadh, Saudi Arabia, in 2002 and the
M.S. degree from the University of Ottawa, Ottawa,
ON, Canada, in 2006. He is currently working to-
ward the Ph.D. degree at the School of Information
Technology and Engineering, University of Ottawa.
In 2003, he was with the Computer Science
Department, College of Computer and Information
Sciences, KSU. He is a member of the Multimedia
Communications Research Laboratory (MCRLab)
and the Distributed and Collaborative Virtual Environments Research Labo-
ratory (Discover Lab). His research interests are in the field of haptics,
computer security, including authentication and authorization, and 3-D virtual
environments.

View publication statsAuthorized licensed use limited to: IEEE Xplore. Downloaded on March 5, 2009 at 02:38 from IEEE Xplore. Restrictions apply.