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

net/publication/229025235

Modeling and Simulation of a Robust e-Voting System

Article · January 2009

CITATIONS READS
22 1,189

8 authors, including:

Mohammad Malkawi Mohammed A Khasawneh

Jordan University of Science and Technology University of Jordan
49 PUBLICATIONS   493 CITATIONS    53 PUBLICATIONS   291 CITATIONS   

SEE PROFILE SEE PROFILE

Omar Al-Jarrah
Jordan University of Science and Technology
69 PUBLICATIONS   1,415 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Pattern Recognition View project

artificial intelligence and image processing View project

All content following this page was uploaded by Omar Al-Jarrah on 17 May 2014.

The user has requested enhancement of the downloaded file.

198 Modeling and Simulation of a Robust e-Voting System

Modeling and Simulation of a Robust e-Voting System

Mohammad Malkawi, Associate Professor, Argosy University, and VP, AIM Wireless USA
[email protected]
1
Mohammed Khasawneh , IEEE Senior Member, College of Engineering, [email protected]
Omar Al-Jarrah2, Associate Professor of Computer Eng., and Laith Barakat2
1
University of Illinois at Urbana Champaign
2
Jordan University of Science & Technology

Abstract cards. These systems are not reliable because of

In this paper we present a simulation model for a problems in reading cards and were replaced by
multifaceted online e-Voting system. The proposed optical scan device systems, which allow voters to
model is capable of handling electronic ballots with record choices by filling in areas on the ballots. The
multiple scopes at the same time, e.g., presidential, ballots are read using a computer scanner and then
municipal, and parliamentary, amongst others. The the votes are counted automatically using a
model caters for integrity of an election process in computer program [1]. Finally, special-purpose
terms of the functional and non-functional computers are used as voting machines where voters
requirements. The functional requirements use touch screens or push buttons to select choices,
embedded in the design of the proposed system which are stored and counted or processed by a
warrant well-secured identification and special program on the same machine [1].
authentication processes for the voter through the
use of combined simple biometrics. Of utmost Often times, however, counting errors take place,
importance are the requirements for correctness, and in some cases, voters find ways to vote more
robustness, coherence, consistency, and security. To than once, introducing irregularities in the final
verify the robustness and reliability of the proposed count results, which could, in rare cases, require a
system, intensive computer simulations were run repeat of the election process altogether! Moreover,
under varying voting environments, viz. voter in some countries, purposely introduced
density, voter inter-arrival times, introduced acts of manipulations of the votes take place to distort the
malice, etc. Results of the simulations show the results of an election in favour of certain candidates.
impact of several parameters on the performance of Although such mishaps can be avoided with a
the system. These results provide the proper properly scrutinized election process, errors can still
grounds that would guide the decision maker in occur, especially when the number of voters is quite
customizing an e-voting system. large. Quite often, international monitoring bodies
are required to monitor elections in certain
Key-Words: e-voting, modeling and simulation, countries.
performance evaluation
The advancement of information and
1. Introduction telecommunications technologies allow for a fully
automated online computerized election process. In
Election is a process in which voters choose their addition to overcoming commonly encountered
representatives and express their preferences for the election pitfalls, electoral vote counts are done in
way that they will be governed. Correctness, real time that by the end of elections day, the results
robustness to fraudulent behaviors, coherence, are automatically out [2, 3]. The election process
consistency, security, and transparency of voting are can be easily enhanced with various features based
all key requirements for the integrity of an election on the demand and requirements of different
process. There is a wide variety of different voting countries around the world. E-Voting is an inter-
systems that are based on traditional paper ballots, disciplinary subject and should be studied together
mechanical devices, or electronic ballots [1]. In a with the experts of different domains, such as
traditional paper ballots, voters choose or mark their software engineering, cryptography, politics, law,
favourite choices on ballots and place them in economics and social sciences. Although many
boxes, which are sealed and officially opened under people have worked on this subject, mostly e-Voting
special conditions to warrant transparency. The is known as a challenging topic in cryptography
ballots are then counted manually, which is a because of the need to achieve voter anonymity, and
tedious process that is subject to human error. With therefore, to ensure his/her privacy [4]. However,
voting via mechanical systems, meanwhile, voters many studies warn against the adoption of e-Voting
make their choices by pulling down on mechanical because of many challenges in software
levers that correspond to their favourite choice of engineering, security, and auditing [1, 4 - 7].
candidates. Each lever has a mechanical counter that
reports the number of votes for that position. These Due to worldwide advancements in computer and
machines are no longer manufactured [1]. On the telecommunication technologies supported by the
other hand, some systems use punch cards where underlying infrastructures, online voting or e-Voting
voters punch holes in computer readable ballot is no longer a North American or Western
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
 Mohammad Malkawi, Mohammed Khasawneh and Omar Al-Jarrah 199

phenomenon. This high tech method of casting a voters and the total number of registered voters in a
ballot has spread far beyond the United States, certain district. The available bandwidth for the
expanding throughout the entire world. E-Voting, communication links is an important factor as well.
along with its benefits and detriments, can now be Note that the bandwidth may vary within the same
found from the developed countries of Europe to the country. For example (in some countries) DSL links
developing countries of Asia and South America. are available only to particular localities, while dial
The introduction of electronic voting has been the up links are used more often in other areas. The
biggest change, for instance, to the Irish electoral message size used for communication between a
system since the establishment of the state over 80 voting station and the central servers has a notable
years ago. E-Voting may soon become a global impact on overall performance. This, in turn, will
reality or a global nightmare [8 - 10]. In 2003, a dictate the type and size of authentication traffic that
new e-Voting system was introduced in Belgium in can be accommodated by the system. Another
two locations to convince citizens that the system important parameter is the architecture of the data
was trustworthy [11]. They introduced a "Ticketing" management system. Here, the performance of the
system where the voter prints and approves a hard system is directly impacted with the use of either a
copy of his/her vote. At the end of the elections, all centralized or distributed approach for data storage,
of the paper votes (tickets) are counted and manipulation and management. A distributed
compared to the electronic result. In this election, approach, however, introduces more challenges for
there was an electronic voting problem reported maintaining accuracy/currency of the voting
where one candidate got 4096 extra votes because of process. These parameters are implemented in a
a technical problem [11]. Besides reliable e-Voting generalized simulation model.
technologies, there is a dire need for international
standards to govern the technology, the software In the simulation model, several metrics are used to
reliability and accuracy, the processes and evaluate overall performance and system behavior.
algorithms deployed within the technology, and the The main metrics will be the voting (simulation)
verification of all hardware, software and protocols time, which represents the ability of the system to
involved. Such standards will eventually allow execute the voting process in the allotted timeline.
elections to proceed in any part of the world without Internally, the average queue length (average
the need for monitoring bodies. The design of a number of voters waiting to vote) is another metric.
“good” voting system, whether electronic or using The average waiting time per voter is yet another
traditional paper ballots or mechanical devices must important metric. One of the key tuning parameters
satisfy a number of sometimes competing criteria would be the number of voting stations required at
including a high degree of security and accuracy, each voting center. This parameter is important,
eligibility and authentication, integrity, verifiability because it, by and large, is the only parameter which
and auditability, reliability, flexibility, performance can be tuned during the voting process (given the
and scalability [12, 13]. availability of hardware resources). The voting
center manager can (in principle) add or remove
More importantly, there is a real need for a good stations as deemed appropriate.
simulation model which can guide the deployment
of e-Voting resources such that the election process The rest of this paper is organized as follows. In the
can proceed with minimal faults and performance next section (section II) we describe in more detail
issues. In this paper, we provide a simulation model the general e-Voting model, while the simulation
for a generic e-Voting process. The model is model is introduced in section III. Simulation results
designed to be flexible enough to be adapted to are presented in section IV. Finally, the paper is
different election environments. The objective of the concluded in section V.
simulation model is to study the effect of several
parameters on the course of an election process. 2. The Proposed e-Voting Model
Simulation results provided by the model, for a
particular election process, allow offices Automating an election process, while relying on
administering an election process to deploy state-of-the-art in computer and ICT technologies,
adequate hardware and networking resources to can significantly mitigate many of the factors that
make the process as successful as possible. would hamper a healthy progress of a given election
process. For automated e-Voting processes to be
There are several parameters which impact any fully acceptable worldwide, several issues must be
voting process. The rate at which voters arrive at addressed and resolved. Among these issues are
voting centers has a direct impact on overall system authentication/validation, security, robustness,
performance. Hence, a heavy arrival rate at a certain performance and correctness. Given the short
voting center may require more voting stations in history of e-Voting systems across the world and the
order to complete the voting process in a timely inherent limitations in the scope of implementation,
manner. The simulation model allows for a good it is very difficult to measure the success or failure
estimation of the number of voting stations at each of any or all of the issues mentioned above. In
center based on a predicted average arrival rate of addition, any voting process, as mentioned earlier, is
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
 200 Modeling and Simulation of a Robust e-Voting System

bound by regulations and cultural values that functional requirements. Of utmost importance are
characterize the different societies involved. Hence, the requirements for correctness, robustness,
the example of one country may not directly suite coherence, consistency, performance and security.
the example of another. As a result, it is highly
recommended to build a simulation model whereby The client side represents a voting station, where
an e-Voting system can be evaluated and various voters cast their votes. Note that the hardware on
attributes adequately assessed before one is the client side includes IO devices for verification
deployed. and authentication (e.g., image scanners, ID card
readers, finger print readers, etc.). In addition to
This paper introduces one simulation model, where that, two more requirements are necessary. In order
we address the main factors which directly to reduce the traffic rate on the network links, a
contribute to the success of a voting process. The local database at the client side is required to host
simulation parameters can be changed based on the the data which pertains to the local voting center.
peculiarities of any entity. The main components of This DB is a rather dynamic one, in the sense that
the architecture of the model are shown in Figure 1. the data stored in its tables may vary over the
This is a client/server web-enabled architecture. election time period. The size of the local DB at any
voting center is only a small fraction of the global
DB at the server side. The use of a local DB

Statistics Statistics
Correctness
Administration
Administration
Correctness Robustness
Local Database
Database
Authentication
Voting Candidates Voters Coherence
Candidates Voters
Process Coherence
OS: Windows/ Consistency Consistency
UNIX/ IODrivers OS: Windows/
UNIX
Hardware: Security
Security
Card Readers
Image Scanner Hardware
Finger Print
Server Side Software System
Client Side SoftwareSystem
Figure 1b
Figure 1a enhances the performance of the voting process.
However, this approach creates a synchronization
The server side hosts the central database for the problem, which will be addressed later in this paper.
voting as well as the candidate population. The The alternative is to use one centralized DB. The
server also collects basic statistics related to an second requirement is the transparency of the voting
ongoing election process (some statistics can be process. In essence, a voter on an electronic voting
turned on or off based on the needs and station casts his/her vote to a computer. The voter
requirements of each election unit). Besides the does not have an insight on how his/her vote is
main functional properties of a voting system, as translated and/or tallied. In a paper-based election,
described in the previous section, the e-Voting the ballot is filled out by the voter and dropped into
system must cater for several essential non- a sealed box by the voter himself/herself. Votes are
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
 Mohammad Malkawi, Mohammed Khasawneh and Omar Al-Jarrah 201

counted in the presence of candidates or their even if the result of the vote is not committed to the
representatives. central DB in due time, the flag in the voter’s
central record is set to TRUE, thus eliminating the
The voter is certain that his/her exact ballot with possibility of another attempted voting by the same
his/her vote selection is placed in the appropriate voter, or by someone who carries a counterfeit ID
box. Of course, ambiguity in the ballot formats (as card. This requires that whenever the record of a
was the case in the US presidential election in 2000) voter is accessed for identification, even when the
may render the transparency a rather deceiving one. record is found at the local DB, the flag on the
In an electronic version, the voter puts his trust into central record must be checked. If it has already
computer hardware, software and network been set to TRUE, the voter is denied access and
infrastructure that process his/her vote. Hence, the his/her attempt fails to go through. If two people
e-Voting system in its broadest form may render the carrying the same ID card (one is authentic while
process a non-transparent one [8, 14]. This issue can the other is counterfeit) attempt to vote
be resolved by printing a hardcopy of each vote for simultaneously, the first one to access the record
the voter to keep for his/her records. Another copy will set the flag to TRUE, load the record and
is printed, possibly in the form of a bar code, and prevent the other one from accessing the record. Of
saved for later verification. In order to verify the course, if the one with the counterfeit card obtains
accuracy and correctness of the process, a random the record first, the vote cast will fail at the next
sample of the saved copies can be scanned and authentication step. It is possible that a record gets
verified against the votes stored in the DB. loaded into two different voting centers due to block
transfer from the central DB into local DB’s. When
The identification of a voter is done via a card a voter attempts to access the record from any of the
reader which reads off his/her official ID card and stations, the client will verify the central record flag.
retrieves the voter record from the local DB (on the If it has been set to TRUE, access is denied;
client side) or loads the record from the central DB otherwise it sets the flag to TRUE and access is
if it is not already in the local one. Records are granted. Note that simultaneous requests to the same
loaded dynamically from the central DB to the local record will be synchronized by the DB query
DB’s either on demand or on a pre-fetch basis. The serialization process (only one query may access
voter record includes, amongst others, a biometric any table at any given time). This mandatory check
description of the voter in question. In this study, we of the flag in the central DB will add extra overhead
use a fingerprint authentication method (other on the network. This overhead is already included in
methods can be added to the model). The voter will evaluating the simulator performance and is
be rejected if his/her fingerprints do not match the reflected into the ensuing simulations.
stored record. In order to reduce false rejections, we
store for each voter several copies of his/her Another synchronization resolution is required when
fingerprints taken at different time periods. a vote is to be tallied into the record of a candidate.
Fingerprints are stored as an encoded text in order to If a candidate is being selected by several voters at
reduce storage consumed by images. This dual the same time, then a certain assignment plan needs
process should guarantee that no voter can falsely to be put in place so that all votes will be tallied (no
impersonate another. Note that the use of misses) and added to the candidate’s record. Again,
fingerprints or any other scanned image directly we use a “COUNT” flag/mutex for the candidate’s
impacts the message size and hence the performance record. The COUNT flag is initially set to FALSE.
of the network. Hence, a distributed database When the record is selected by a voter, the flag is set
approach is preferable over a centralized approach. to TRUE until the record count is updated, then the
flag is reset to FALSE. All votes for the same
The accuracy and correctness of the e-Voting candidate will be queued until the flag is reset to
process can be further jeopardized if the same voter FALSE. In order to improve the ‘hit’ performance, a
casts two or more votes, or a vote is not properly
added to the overall count of the right candidate.
Such mishaps may come about as a result of
synchronization conflicts at the central DB level. In
order to prevent two or more votes per voter, we use
a “voting status flag” in the voter record. This flag is
initialized to FALSE. The voting status flag is set to
TRUE in the central DB whenever a voter identity is
verified (before authentication takes place). If the
authentication fails, the flag is reset to FALSE. If
the voter leaves the station without completing a
vote, the flag is also reset to FALSE; thus allowing
the voter another chance to try again and cast his/her
vote. If the voter successfully completes the voting
process, the flag remains set to TRUE. Note that
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
Figure 2: Voting process flow chart
202 Modeling and Simulation of a Robust e-Voting System

counting semaphore COUNT can be used instead. A

copy of the vote will be printed only when the vote
is successful and the candidate’s record is updated.
This requirement, initially made for transparency
purposes, provides a final test for the accuracy and
correctness of the process, especially in the presence
of thread hang-ups. Figure 2 shows a flowchart of
the voting model.

The overall architecture of the system is shown in

Figure 3. The central database in Figure 3, which is
mirrored out for reliability reasons, is used to store
all relevant information on the candidates and
voters. Voting centers are distributed around the Figure 3: G eneral vie w of system architecture
given country. One or more voting centers could
share a local database. A voting center consists of high voter densities, respectively. For example, λ=5
one local DB server and one or more voting stations. and µ=10 represents the case when voters arrive in
large clusters but at rather dispersed time intervals.
Information security is very important to our The number of voters expected to vote at a given
system. There is inherent need to secure all the voting center is predefined. The model simulates
communications between the clients and their local centers with as few as 2000 voters and centers with
DB servers. We also need to secure the as many as 20,000 voters. In general, the simulator
communications between the local DB(s) and the model allows for a random number of voters to be
central DB server [15]. Security concerns and selected per center. The simulator is expected to run
solutions will be presented in a separate work by the until all voters registered at each center have cast
authors. their votes. Simulation is normally set to complete
within 12 hours (typically voting begins at 7:00 am
Each voting station is equipped with a card reader, a and completes at 7:00 pm). However, the simulator
fingerprint scanner, a touch screen, and a can be tuned for any required simulation time
multimedia subsystem. The multimedia subsystem period.
is used for people with special needs (physically
challenged), such as the blind and those with Voters are queued at the voting stations within each
difficulties in reading or comprehending images, voting center. A voting center consists of (N) voting
texts, or sounds. Figure 3 also shows the structure of
a voting station.

The proposed system is capable of handling

electronic ballots with multiple scopes at the same
time, e.g. presidential, municipal, parliamentary,
and others. However, the simulation environment in
this study is designed only for a single voting scope.

3. Simulation Study of the Proposed Model

The simulation model is a general (M/S/G) queuing

model. Voters randomly arrive at a voting center
according to a Poisson random process. The inter-
arrival rate is controlled by the mean time between
two successive arrivals (mean interarrival time,µ)
and is governed by an exponential distribution. The
simulator allows for as many voting centers as
needed. The rate of voter arrival varies over time;
low arrival rates characterize early hours of the day; Figure 4: simulation environment
heavy arrival rates characterize mid morning times
and close to elections closing time. In the simulator, stations, and hence N queues. When a group of
we choose µ =10, 5 and 2 for low, moderate and voters arrive, the simulator moves the voters to the
large inter-arrival rates, respectively. appropriate queue either in a round robin manner
(one voter per queue at a time) or enters the whole
Voters arrive at a voting center in a rather clustered cluster of voters into the least loaded queue. Both
manner; i. e., in groups. The average size of a scheduling policies are evaluated in the simulator.
cluster is a Poisson random variable with mean (λ). Other policies can be added and evaluated.
In our simulations, we use λ=2 and λ=5 for low and
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
 Mohammad Malkawi, Mohammed Khasawneh and Omar Al-Jarrah 203

The service time (i.e., the time it takes to complete voting runs. This is particularly important because
one whole voting transaction) is a blend of real time the model entails several random factors.
execution and random delay due to several factors. The simulation environment entails a live Oracle
The random delay is made up of the average time database system for voters and candidates. Besides
required to read the voter’s ID and scan the personal identification information, the records
fingerprint when a voter is de-queued and selected include authentication information and locality of a
to vote. This average is empirically determined voter and/or a candidate. The simulator, also,
using typical card readers, scanners and touch includes modules which emulate the arrival of
screen monitors. The verification and voters at voting centers and the voting process itself.
authentication processing time consists of real-time The simulator allows a voter to cast a vote at any
access to an Oracle database. DB transactions voting center, irrespective of his actual voting
undertaken in the simulator include setting the voter district (locality). This is one of the main advantages
flag as well as the candidate Counter semaphores. of an online e-Voting system.

In the simulation model, as figure 4 shows, the We have conducted a fairly large number of
central DB server and the local voting station simulations of the proposed voting system, taking
servers are located within the same local network the number of voters over a sample range starting at
segment. Therefore, we introduce a random delay to 2000 voters per voting center and ending at 20,000
compensate for inter-net transfer time. The transfer voters per voting center. We realize that the number
time is a function of the available network of voters in a given locality may be much larger
bandwidth and message size. We use several than the numbers we used in the simulator.
bandwidth sizes in this study. A 1 Mbps is a rather However, the simulation results are fairly scalable
conservative bandwidth and is typical of many where the simulation model is capable of modeling
voting locations around the world. fairly large number of voters. We fixed the number
of voters at a given voting center in the simulator.
Although in reality, this number may vary by a
small percentage due to the fact that people will be
allowed to vote at any other center they choose for
the sake of voting convenience, especially those
voters residing at townships outside their voting
districts, or those voters casting their votes through
embassies away from their home country/ies.

4. Simulation Results:

In the following subsections we discuss the impact

of various model parameters on the overall model
and its performance.

4.1. Centralized Versus Distributed DB

The architecture of the proposed e-Voting model

can implement either a centralized or distributed DB
approach as discussed earlier. The centralized
Figure 5: State Diagram of the Proposed approach keeps all the records of the voters and the
Simulation Model candidates in a central DB located at a central
server. Each voting transaction must interact with
We also use a smaller bandwidth (128 Kbps) which the central DB. In a distributed approach, however,
is typical of dialup locations. Furthermore, we use a local servers at voting centers download voter and
10 Mbps bandwidth for more developed network candidate records most relevant to the local center
infrastructures. The simulator uses a Weibull on demand or on a pre-fetch basis. In the
distribution to model the bandwidth which accounts centralized approach, data consistency is not a
for delays in the network. The message size is a serious issue since all data is maintained at one
function of the voter and candidate records. The central DB. However, this approach causes a serious
simulator runs for several message sizes. Note that performance problem both in DB response time and
the use of biometric measures for verification and network traffic. In the simulated model, we
authentication produces larger message sizes and, implement a quasi-distributed approach. A central
hence, increases the overall transfer time of data DB is used to host all candidate and voter data. Each
across the network. voting center has a local DB server which loads the
candidates as well as the voters registered at the
The voting process is shown in the state diagram of local voting center. The data consistency issue and
Figure (5). The simulations were averaged over five the accuracy of voting are maintained by means of
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
 204 Modeling and Simulation of a Robust e-Voting System

synchronization flags as discussed earlier (the This allows the administering body in a certain
VOTING and STATUS flags shown in Figure 5). country to properly size a given voting center given
When the flags were turned off we noticed several the number of voting populace and the available
violations of the voting accuracy. In the worst case bandwidth. Figure (7) shows the time required to
scenario, we noticed 1.2% error rate; where the error complete all voting transactions given a certain
is manifested in a cast vote not being reported in the number of voting stations and a voting population.
final results for a given candidate. With the use of This figure illustrates the trade-off between the
the synchronization flags, errors of the like are voting time and the number of voting stations. In
totally eliminated. countries, where the cost of voting stations can be a
real burden, the voting time period can be extended.
4.2. Number of Voting Stations For example, 20000 voters can be served by 2
When the number of voters at a given center is stations over 35 hours or can be served in 12 hours
relatively large, or the arrival rate of voters is high, using 36 voting stations.
it is recommended to add more voting stations to the
center. In a centralized approach, the addition of 4.3. Network Bandwidth
voting stations may not improve the results The wide area network bandwidth has a direct
significantly. impact on the model performance. Figure (8) shows
the average service time for three different
40
bandwidth values (128 Kb/s, 1 Mb/s, and 10 Mb/s).
35
Impact of Network Bandwidth
30

12
25
Number of VotingStations

10
20

15 8
AverageServiceTime(s)

10 M

10 6 1M
128 K
5 4

0
2
0 5000 10000 15000 20000 25000
Number of Voters
0
1 2 3 4 5 6 7 8 9 10 11 12
Figure 6: Voting Center Configuration VotingstationIDNumber

More voting stations simply shift the overload from

the local station on to the main (central) server. Figure 8
However, in a distributed approach, the addition of
local stations will distribute the load locally without We show the results for different voting stations;
significantly impacting the main server. here, we show the network effect at the various

Figures (6) shows the number of voting stations 60.0

required to complete all voting transactions within 50.0
12 hours for different voting populations. For
example, 36 voting stations are needed to allow all 40.0
128 K
AverageQueueLength

20000 voters to vote within 12 hours (7:00 am to

30.0 1M
7:00 pm). The results shown in the figure assume a
1 Mb/s bandwidth and a clustered policy in a 10 M
20.0
distributed architecture.
10.0
Distribution of Voting Time Over No. of Stations
40 0.0

35
1 2 3 4 5 6 7 8 9 10 11 12

30 Voting stationID Number

8000
Total Voting Time

25 12000

20 16000
Figure 9: Network Bandwidth
20000
15

10
voting stations (in this case 12 stations) being
incubated at a particular voting center. The service
5
2 6 10 14 18 22 26 30 34 38
time is defined as the time it takes to process a vote
No. of Voting Stations Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
F igu re 7
 Mohammad Malkawi, Mohammed Khasawneh and Omar Al-Jarrah 205

once the voter has been admitted to the system. This Voters arrive at a station at an average rate of (1/ µ)
time includes identity verification, authentication, according to an exponential inter arrival rate.
flag checks, the update to the central DB and the Furthermore, voters arrive in clusters (groups at a
statistics update as shown in Figure 8. Beyond a time) with a mean (λ) according to a Poisson
bandwidth of 1 Mb, any pronounced improvement random process. The scheduling policy used in this
begins to diminish. The x-axis represents the voting study (clustered) allocates all (λ) voters to the
station ID used to process the voter population. The station with the shortest queue length. We observed
results in Figure 8 are averaged over a 12-hour that this scheduling policy, although intuitive, leads
simulation timeframe for 10,000 voters. to the fact that some stations may become free,
while others remain busy, especially with large
Furthermore, the simulator was tuned to run and number of voting stations in the voting center. For
accommodate all voters within a 12-hour timeframe. comparison purposes, we implemented another
Figure (9) shows the average queue length for policy (round robin) which aims at keeping a
different bandwidths. The shorter the queue is the balance between the various voting stations. This
less time a voter expects to wait before casting a policy allocates voters in a round robin manner
vote at a voting station. The average queue length across all voting stations. Figures (10a and 10b)
can be 3 times as big for low speed network show the average queue length at 5 voting stations
connections (128 Kb/s). Figure (9) also confirms using clustered and round robin scheduling policies,
that the performance improvement beyond 1 Mb/s is respectively. Note that the round robin policy
rather insignificant. maintains a fairly good balance of voters among all
queues. The clustered policy shows a wide variation
among the voting stations in terms of the queue
4.4. Scheduling Policy lengths and the completion time. Some of the
The scheduling of voters to the available voting stations were observed to finish up all the voters
stations at a given voting center has a direct impact within 10 hours, while others persisted for more
than 12 hours. The main advantage of the clustered
policy is the simplicity of implementation and low
cost, since it does not require any distribution of
Queue 1 voters among stations; whereas the round robin
120
Queue 2 requires more personnel to manage the voting
Queue Length

100
Queue 3 process.
80
60 Queue 4 5. Conclusions
40
20
Queue 5
0
In this paper, we have proposed an online e-Voting
system which can tackle all earlier issues
0.6
1.87
2.92
3.98
5.13
6.23
7.29
10.6

encountered in a conventional (manual) voting

Time(H) system. The new system maintains voting statistics
Figure 10a: Clustured Scheduling in real-time while preserving the integrity of the
voting process from the minute a voter steps in to
cast his/her vote until the cast vote is registered in
favour of the chosen candidate at a globally
allocated DB repository. While observing full-
fledged voting transparency, at the voter as well as
the system levels, the proposed system is capable of
18 denying access to any illegal voter/s, preventing
16 Queue 1 multiple votes by the same voter, and blocking any
Queue Length

14
Queue 2 introduced forms of malice that would adversely
12
10 Queue 3 affect the voting process altogether. Moreover, the
8 proposed voting system caters for the needs of the
Queue 4
6 physically challenged voters by providing special
4
Queue 5 multimedia amenities that would facilitate voting to
2
0
a voter’s convenience.
0.6
1.87
2.92
3.96
5.1
6.19
7.24

10.2

Simulation results of the system, while running a

Voting Tim e (H) live DB backend server, reveal a number of
Figure 10b: Round Robin Scheduling
important factors that ought to be assessed carefully
by the party adopting a system like this one, for any
form of election activities, prior to its final
deployment. These factors address the number of
on the overall performance of the voting process. voting stations needed at any voting center, as
outlined by the voting needs of a given voting
Communications of the IBIMA
Volume 8, 2009 ISSN: 1943-7765
 206 Modeling and Simulation of a Robust e-Voting System

district, the network bandwidth requirement by a “Recommendations for Congressional

given voting center, and the size of the local DB to Action,” August 2001
support the needs of a given voting locality, 7. California Internet Voting Task Force. A
amongst others. The system, via these simulations, Report on the Feasibility of Internet
has shown ruggedness and sustained reliability in Voting, Jan. 2000.
terms of preventing multiple votes by the same http://www.ss.ca.gov/executive/ivote/.
voter, and maintaining internal system audits that 8. McGaley Margaret, McCarthy Joe,
would warrant no missed votes, per candidate, in the “Transparency and eVoting: Democratic
process of voting. vs. commercial interests”, 2004.
www.cs.nuim.ie/~mmcgaley/Download/Tr
With the use of an e-Voting system, as the one
ansparency.pdf
proposed in this paper, many of the issues, that had
9. Online Voting. Parliamentary Office of
long challenged traditional voting systems, are
bound to be resolved providing a peace of mind to Science and Technology. May 2001.
both voters and election candidates. It is well www.parliament.uk/post/pn155.pdf
expected that with a well administered/designed e- 10. McGaley, Margaret. “Irish Citizens for
Voting system, countries that have long been Trustworthy Voting.” 6 July 2004.
observed by international monitoring bodies, while http://evoting.cs.may.ie/
carrying out election processes of their own, will 11. "Electronic voting in Belgium",[Online],
soon be able to work on their own and, yet, achieve Available:
the election integrity they have longed for. http://en.wikipedia.org/wiki/Electronic_vot
ing_in_Belgium
Issues of security and reliability at the central DB, 12. "The Problem with Electronic Voting
the communication links and local servers will be Machines", 2004. [Online], Available:
further addressed by the authors in a separate study. http://www.schneier.com/blog/archives/20
04/11/the_problem_wit.html
References 13. "Development of remotely secure e-voting
system" Keshk, A.E.; Abdul-Kader, H.M.;
1. National Science Foundation. Report on Information and Communications
the National Workshop on Internet Voting: Technology, 2007.
Issues and Research Agenda, Mar. 2001. 14. TADAYOSHI KOHNO, ADAM
http://news.findlaw.com/cnn/docs/voting/n STUBBLEFIELD, AVIEL D. RUBIN,
sfe-voterprt.pdf DAN S. WALLACH: Analysis of an
2. R. Mercuri. Electronic Vote Tabulation Electronic Voting System, IEEE
Checks and Balances. PhD thesis, Symposium on Security and Privacy 2004
University of Pennsylvania, Philadelphia, Introduction to Computer Security, by Matt
PA, October 2000. Bishop (ISBN: 0-201-44099-7), Addison-
3. A. D. Rubin. Security considerations for Wesley 2005.
remote electronic voting. Communications
of the ACM, 45(12):39–44, December Copyright © 2009 by the International Business
2002. http://avirubin.com/e- Information Management Association (IBIMA). All
voting.security.html rights reserved. Authors retain copyright for their
4. Cetinkaya, O. and Cetinkaya, D. (2007) manuscripts and provide this journal with a
“Verification and Validation Issues in publication permission agreement as a part of
Electronic Voting” The Electronic Journal IBIMA copyright agreement. IBIMA may not
of e-Government Volume 5 Issue 2, pp 117 necessarily agree with the content of the
- 126, available online at www.ejeg.com manuscript. The content and proofreading of this
5. "How E-voting Works", [Online], manuscript as well as and any errors are the sole
Available: responsibility of its author(s). No part or all of this
http://people.howstuffworks.com/e- work should be copied or reproduced in digital,
voting.htm/printable hard, or any other format for commercial use
6. Caltech/MIT Voting Technology Project, without written permission. To purchase reprints of
this article please e-mail: [email protected].
“What Is, What Could Be,” July 2001,
http://www.vote.caltech.edu/Reports/2001r
eport.html ; Carter et al, To Assure Pride
and Confidence; The Constitution Project’s
Forum on Elections, “Building Consensus
for Election Reform,” and

Communications of the IBIMA

Volume 8, 2009 ISSN: 1943-7765

View publication stats