WATA
A System for Written Authenticated though Anonymous Exams
Giampaolo Bella, Gianpiero Costantino and Salvatore Riccobene
Dipartimento di Matematica e Informatica, Universit
`
a di Catania, Catania, Italy
Keywords:
Exam, Open Competition, Authentication, Anonymity, Platform independence.
Abstract:
University exams or open competitions raise significant security concerns because they are typically driven by
strong interests. The candidate and the examiner may want to cheat on each other, or even coerce each other.
While candidate authentication would address certain threats, his anonymity would address some other threats,
but their conjugation appears to be contradictory. The classical approach to facing this security problem is
weak as it poses significant trust upon the examiner. It is the double envelope, with a big envelope containing
an anonymous exam sheet and a smaller envelope in turn sealing the candidate’s identity. Surprisingly, there
appears to be almost no computer-assisted solutions available.
WATA is a system, implemented in a portable software, for Written Authenticated though Anonymous exams.
It protects the candidate and the examiner from each other by ensuring that the examiner corrects an authenti-
cated though anonymous exam sheet. The candidate is the sole entity who can establish the link between his
identity and his exam sheet, and normally has interest in safeguarding such a link. With various additional
functionalities, WATA can be freely downloaded from the Internet for academics to try (Gianpiero, 2009).
1 INTRODUCTION
Taking an exam typically is a sensitive step for a can-
didate. Success may signify advancing towards a uni-
versity degree or getting an important job. It is there-
fore an obvious requirement that both the candidate
and the examiner behave honestly in their respective
roles.
Nevertheless, various forms of cheating may oc-
cur. For example, the candidate may try to send some-
one else, who is particularly skilled and thus certain to
pass, to sit for the exam on his behalf. In practice, the
accomplice might be a friend or even an expert who
is paid for his illegitimate task. Authentication can ef-
fectively withstand these threats, and in this context
can be read as: the candidate really has the identity he
claims on his exam sheet. The obvious enforcement
of authentication during an exam is to have the exam-
iner check that the candidate’s ID matches what the
candidate writes down on his exam sheet.
However, other forms of cheating are possible dur-
ing an exam. The candidate might explicitly or im-
plicitly influence an examiner who wants to mark the
exam sheets fairly but happens to be a friend or rel-
ative of the candidate’s. The candidate might even
coerce the examiner. Also the examiner may want to
act dishonestly by favouring or disfavouring an exam
sheet for whatever reason, for example because he
likes or dislikes the corresponding candidate. Re-
markably, these forms of cheating are exacerbated by
authentication, because the examiner is entitled to as-
sociate each exam sheet to a candidate. Anonymity
can effectively withstand these threats, and in this
context can be read as: the candidate can effectively
conceal his identity on the exam sheet. It is clear that
authentication and anonymity are, at least at the ab-
stract level, the opposite of one another. Therefore, a
security system to thwart cheating during an exam is
not trivial to design because it cannot simply enforce
one of the two goals, while enforcing both of them
appears to be conceptually impossible.
The original motivation to tackle this problem
dates back to 2003 for the inception of the Computer
Security course towards the degree in Computer Sci-
ence at our university. Not only was it was felt that a
computer-assisted solution to the problem would pro-
132
Bella G., Costantino G. and Riccobene S. (2010).
WATA - A System for Written Authenticated though Anonymous Exams.
In Proceedings of the 2nd International Conference on Computer Supported Education, pages 132-137
DOI: 10.5220/0002797401320137
Copyright
c
SciTePress
vide fairness to both the examiner and the candidate,
but it was also taken as a stimulating research chal-
lenge. In fact, it was somewhat surprising to find out
that no attempts existed at tackling the problem using
computers, as there was no public availability of the
only related publication (Gray, 2003), which was just
about to appear (§2). Our efforts to study the prob-
lem from a security perspective produced the first beta
version of WATA, a system for Written Authenticated
Though Anonymous exams, which has recently met a
stable Java implementation, as we shall see below.
The gist of WATA is to mechanise in a software
the classical method of the double envelope, which is
often used during open competitions. As we detailed
below, because this system classically has no com-
puter support, it is easy for a dishonest examiner to
subvert it. The main idea is to authenticate each exam
sheet conventionally, but to keep it anonymous for the
examiner until he terminates his marking. The only
real limitation of WATA is against a scenario where
the candidate and the examiner collude to subvert the
exam. This seems, however, rather extreme, and may
as such require dedicated enforcement.
This manuscript begins with a survey of the re-
lated work (§2). It then describes WATA (§3), illus-
trates its interface (§4) and outlines its implementa-
tion (§5). Finally, it terminates (§6).
2 RELATED WORK
As mentioned above, only two significant works are
related to ours at present. One is the double paper
envelope, which is a classical attempt at conjugating
authentication and anonymity without computer sup-
port. Two envelopes are used for each candidate. A
big one, which must be anonymous, will contain his
exam sheet. A small one will contain only the can-
didate personal information. When the candidate fin-
ishes his exercises, the examiner puts the exam sheet
inside the big envelop, whereas his personal informa-
tion gets sealed in the small envelop sealing-wax
could be used to make the small envelop more tamper-
proof. During the marking phase, the examiner will
open only the big envelop in order to mark an anony-
mous exam sheet. Only afterwards, when the exam-
iner will have marked the exam sheet, he will asso-
ciate the anonymous sheet with the personal informa-
tion of the candidate by opening the small envelop.
This terminates the marking phase, which is authenti-
cated though anonymous only if the examiner is hon-
est.
Clearly, a dishonest examiner could easily violate
the anonymity of the exam sheet. If he is not super-
vised by someone else, or if all members of the exam-
ining committee agree, he/they can decide to open the
small envelope beforehand, and dishonestly advan-
tage or disadvantage some candidate. Various cheat-
ing scenarios could take place, especially if the small
envelop is weakly sealed, because anyone in the ex-
amining committee could open and close it without
any apparent tampering. As mentioned, sealing-wax
would be of some help here.
Another related work is a software to sub-
mit coursework in an electronically from the stu-
dent to the examiner while maintaining the student
anonymous (Gray, 2003). Based on the client-
server paradigm, the system lets the student use a
pseudonym to submit his coursework. This process
is mechanised through a JAVA (Microsystems, 1991)
application, which keeps the identity of the student
from the examiner. The latter is therefore deemed to
mark the coursework fairly. Only when the course
(that is, the actual teaching classes) terminates, will
the examiner associate the pseudonym with the real
identity. The system has weaknesses, as the author
himself says; for example, a student could submit
his coursework with his personal information in or-
der to invalidate the anonymity of his work. This may
happen either deliberately or indeliberately, that is by
mistake. It remains questionable, however, whether
this qualifies as a significant weakness in a realistic
threat model where the student has interest in con-
forming to the rules in order to get through the exam.
The goals of this system are rather different from
WATAs. Rather than assisting with coursework sub-
mission prior to the end of teaching, when anonymity
is relaxed, WATA is targeted at exams whose marks
must be produced in due time, as is the case of open
competitions or final exams of university courses, for
example. Another difference is that WATA is a tool
for the examiner and, as such, does not need to be
distributed.
3 OUR SOLUTION: WATA
In this Section, we advance a system called WATA. It
is a system for Written Authenticated Though Anony-
mous exams, which offers an examiner full computer
support to print, mark and notify the marks of exam
sheets that are anonymous though still authenticated.
Anonymity is relaxed only at the final phase of mark
notification, that is after the actual marking phase.
WATA is an application software that accomplishes its
goals by printing exam sheets with a particular layout.
WATA - A System for Written Authenticated though Anonymous Exams
133
3.1 Authentication
To support authentication, we design what we name
an authentication token (Fig.1), which essentially is a
form for personal information.
Figure 1: Authentication token.
WATA prints this token on the exam sheet. At the
beginning of the exam, the candidate is required to fill
it in with his personal information (name, surname,
student enrollment number), date and signature.
Figure 2: Top of an exam sheet.
The barcode is the most important feature of the
token. Because it is unique, and is duplicated on the
exam sheet, it associates the token to its sheet, as we
shall see below (§3.2).
When the candidate has filled in the token, the ex-
aminer authenticates him by asking his identity doc-
ument. If the personal information as shown on the
candidate’s ID matches what he has written on the to-
ken, the examiner puts a signature on the token to val-
idate it. By doing so, he certifies the association be-
tween the candidate and his token. In addition, each
token will be reinforced with a stamp of the exam-
iner’s organization in order to make it more tamper-
proof. The examiner’s signature and the organiza-
tion’s stamp are the only non-digital expedients that
WATA adopts.
Finally, the candidate is required to cut out the to-
ken from the exam sheet and to keep it with him in a
safe place. After this, the candidate, who is authenti-
cated, can start answering the questions on an anony-
mous exam sheet.
3.2 The Barcode
The barcode is printed on the sheet in two different
places, one is inside the token and the other one is
on the top-right corner of the sheet. It therefore is the
only connection between an authentication token, that
is, between an identity, and an exam sheet. Of course,
no barcode must be printed on more than one exam
sheet, otherwise it would become impossible to later
associate the candidate to his exam sheet.
WATA supports the reading of a barcode through a
standard barcode scanner, so all operations described
in the following can be significantly mechanised and
speeded up. It is understood that this raises some trust
in the examiner, who may in practice, at authentica-
tion time, memorise the barcode number correspond-
ing to the candidate he wants to mark disonestly.
To guarantee uniqueness and unpredictability of
a barcode, we use a random function that is detailed
below (§4.1). During the computation of the number
to represent as a barcode, we apply an extra sanity
check that the last extracted number has not already
been computed. This helps to circumvent any weak-
nesses of the implemented random function. Since
the extracted numbers are stored in a database with
the primary key in the number of the barcode, check-
ing whether a number is present is not heavy.
When the generated barcodes are enough for the
exam sheets, the system is ready to print out the bar-
code using the standard barcode codification cod.39.
This codification covers uppercase letters, digits and
special characters such as % but, for the sake of sim-
plicity, WATA only encodes numbers as barcodes.
3.3 Anonymity
An exam sheet is anonymous when an examiner can-
not go back during the marking phase to the candidate
who completed that exam sheet. The authentication
token of WATA declares the owner of the sheet, but
without it the sheet remains anonymous. As explained
before, the candidate must remove the token from the
exam sheet after the examiner validates the token, but
prior to submitting the sheet. In our practical use of
WATA, the examiner enforced this by declaring the
policy that failing to detach the authentication token
would automatically invalidate the exam sheet: hardly
anyone walked away without his token.
Fig.3 shows the exam sheet without the token.
This is what the examiner sees while he is marking
the exam sheet. It can be noticed that the sheet is
anonymous.
CSEDU 2010 - 2nd International Conference on Computer Supported Education
134
Figure 3: Exam sheet without Authentication Token.
3.4 The Database
The structure of the database underlying WATA is
simple, as it consists of three tables: the barcode table,
the question table and the mark table. The barcode ta-
ble is very simple as it stores the generated barcodes.
We shall see that it is particularly useful during the
notification of marks to the candidates in order to de-
tect fake barcodes the candidate might have built by
himself (§4.3). The second table stores the questions
for the exam sheets (Table 1).
Table 1: A question table.
QID Question
1 How did he/she discover the hot water?
2 How far is the sun?
3 What was the best thing before sliced bread?
4 Can crop circles be square?
Each question is associated to a progressive num-
ber called “QID” (acronym of question identifier),
while the questions belong to the “Question” field. In
addition, WATA allows one to manage questions for
different subjects, such as different university courses.
The examiner merely needs to create one such table
for each subject. Creating a table does not require par-
ticular skill; it can be done using the dedicated button
that WATA provides (§4.4).
Table 2: A mark table.
S-Barcode Mark S-ID Date ENRL
5070920 Passed 1 2009/07/09 Empty
6483021 Failed 2 2009/07/09 Empty
1111220 Passed 3 2009/07/09 Empty
3628103 Passed 4 2009/07/09 Empty
The database also contains a mark table (Table
2). The examiner marks an exam sheet that is token-
less. So, he only stores the barcode of the sheet in
“S-Barcode”, the final mark (in “Mark”) and the date
of the exam in “Date”. The examiner has no means to
know the enrollment number “ENRL”, which is only
filled in later, during the notification phase (§4.3).
Therefore, the mark table supports full auditing of the
candidate’s attempts at passing an exam by recording
the mark for each date the candidate took an exam.
4 USING WATA
Using WATA is simplified by a friendly user interface
(Fig.4). The main features can be accessed through
the buttons on the left hand side. The remaining but-
tons interact with the underlying database.
Figure 4: Interface to WATA 2.0.
The Add Questions” button initiates a straightfor-
ward interaction to expand the database of questions
for an exam. The other buttons are explained in the
sequel of this Section.
4.1 Print Exam Sheets
To prevent collaboration between candidates during
an exam, WATA supports the printing of different
questions for each exam sheet. Prior to printing, the
questions are shuffled using a random function, which
is the same as that used for barcode generation (§3.2).
The random function combines the number of sec-
onds elapsed from 1970-01-01 with the output of the
standard random function provided by the program-
ming language. The details can be observed from the
omissis of the source code, which is detailed sepa-
rately (§5.1).
Finally, the shuffled questions form a list to print
questions from. The examiner can choose how many
exam sheets and how many questions per exam sheet
to print. In case the number of available questions is
insufficient, WATA shuffles the questions again. This
feature was added in order to allow an examining
committee to print the desired number of exam sheets
even if the number of questions available were rather
small. Clearly, this would cause lower entropy.
WATA - A System for Written Authenticated though Anonymous Exams
135
Figure 5: Omissis of extraction of shuffled questions from a question table.
4.2 Mark Exam Sheets
When the examiner has finished to mark all exam
sheets, he is ready to insert the marks in the system.
Using the “Mark Exam Sheets” button, the examiner
fills in the fields “S-Barcode”, “Mark”, “Date” in-
side the mark table seen above (Table 2). This is the
only information that the examiner can gather from
an anonymous exam sheet. The enrollment number
(“ENRL”) will be filled in later.
4.3 Notify Exam Sheets
When the examiner has finished marking the anony-
mous exam sheets and stored the marks in the sys-
tem, he is ready to notify the marks to the candidates.
However, he cannot link each exam sheet to its owner
yet. This can be done only through the authentication
tokens. So, the examiner asks each candidate to hand
his token in. As the examiner scans in a barcode, the
system matches it against two tables.
First, the live barcode is searched in the barcode
table to prevent potential attempts at submitting a fake
token with a fake barcode. Without this check, the
candidate might have downloaded our system (Gi-
anpiero, 2009), printed out an exam sheet with the
proper layout, a barcode and some plausible questions
at home, and answered the questions before the exam
date. At the exam, he would have needed to swap the
real exam sheet with his fake, terminated, one before
being authenticated.
If this first check succeeds, the live barcode is
searched in the mark table. When found, the examiner
fills its “ENRL field with the enrollment number read
from the token. This is how WATA audits each can-
didate’s attempts at passing an exam. Then, WATA
opens up a popup window with the candidate’s latest
mark and full audit trail. The examiner can then tell
the candidate his mark if the trail complies with the
exam policy for example, certain university poli-
cies prevent a candidate who fails an exam to try it
again at the very next available exam date.
This process shows how the system guarantees
anonymity of the candidate until the candidate him-
self gives the token back. As noted above, this feature
protects both the candidate and the examiner from
each other, but cannot help towards their collusion.
4.4 Other Features of WATA
WATA also implements the following useful facilities
(Fig.4) to interact with the underlying database.
Show Table: it visualizes the contents of a table,
such as a question table or a mark table.
Create Table: it creates an empty table, for exam-
ple to support exams for more than one discipline.
Rename Table: it edits a table name conveniently.
Import from File: it imports questions from a for-
matted text file, such as those exported from Mi-
crosoft Access.
5 IMPLEMENTATION OF WATA
Portability is one of the most important requirements
for a software. It should be possible to install the soft-
ware on different computers regardless of the Operat-
ing System. To meet this requirement, we decided to
write WATA in JAVA, which allows one to develop
multiplatform-software and to run it through the Java
Virtual Machine (JVM).
Moreover, WATA uses MySQL (Sun-
Microsystem, 1996) as DataBase Management
CSEDU 2010 - 2nd International Conference on Computer Supported Education
136
System (DMBS). The implementation of WATA
is lightweight and can therefore run on all current
machines. The only requirements are a JAVA version
(al least 1.4.2), and a MYSQL version (at least 5.0).
5.1 A Glimpse at the Code
While the full code of WATA can be obtained from
the authors on demand, it may be useful to look more
closely at a key fragment here.
Fig.5 shows the kernel of the method used to shuf-
fle the questions from the question table (§4.1). In
particular, rows 440 to 446 pertain to the SQL query,
which shuffles the questions according to a random
number taken as a parameter. It is interesting to ob-
serve how that random number is computed, as it is
also used for barcode generation (§3.2). Our initial
attempt to compute the random number was to use the
number of seconds elapsed from the date 1970-01-01,
which is output by UNIX TIMESTAMP(). However,
it was soon realised that this is insufficient because the
query could be launched more than once in the same
second and would then yield the same output. So, the
output of the Math.random() function is added.
Then, the loop from rows 451 to 455 extracts the
questions just shuffled and stores them in a dynamic
vector. During the printing phase, WATA takes the
questions from that vector and sends them out to the
printer.
5.2 History and Future of WATA
The version of WATA described in this paper is WATA
2.0, which supersedes WATA 1.0.
WATA 1.0 was the first stable release. It was only
available for Microsoft Windows (Microsoft, 1985),
as it was written in Visual Basic and the database
support was provided by Microsoft Access. The big
drawback of WATA 1.0 was scalability. In fact, when
the examiner tried to print a high number of exam
sheets, the software became very slow because the al-
gorithm used to shuffle the questions was inefficient.
Today, with WATA 2.0, we overcome the scalability
and portability limitations of WATA 1.0
As future extensions WATA, could introduce a
functionality to separate the questions into difficulty
classes. This would enable the system to let the ex-
aminer somewhat control the difficulty of the ques-
tions printed out on the exam sheets. For example, it
would be possible to implement a requirement that
each exam sheet contain one question of difficulty
“medium” and two questions of difficulty “high”.
Another feature that we plan to introduce in the
future is some L
A
T
E
X(LaTeX, 1985) support. The cur-
rent version of WATA allows to store only text ques-
tions, and so no mathematical symbols are allowed. A
L
A
T
E
Xplugin would offer full support to a large range
of symbols.
6 CONCLUSIONS
University exams or public competitions may suf-
fer various forms of cheating. Authentication and
anonymity may alternatively help face some of them,
but are inherently difficult to use jointly.
We advance a solution called WATA to face these
problems in case of written tests. WATA supports the
establishment of exam sessions that are authenticated
though anonymous. The system relies on an authenti-
cation token, which contains the candidate’s personal
information, and a truly anonymous exam sheet. The
authentication token is retained by the candidate who
therefore is the sole entity who can protect his own
anonymity. The token is mechanically linked to the
right exam sheet by scanning a barcode.
WATA has been used in support of the final exam
of the Computer Security course at our university
ever since 2004. The results are encouraging: the
overhead on the lecturer was only negligible, while
the students gladly accepted the new system. WATA
was described at the beginning of the course. At the
end of the course some students felt that WATA had
given them more freedom in their interaction with the
lecturer, because it had dissipated their concerns to
somewhat negatively influence the lecturer’s evalua-
tion of their exams.
The latest version of WATA has been made pub-
licly available through the Internet these days (Gian-
piero, 2009). WATA currently comes under a Creative
Commons (Creative-Commons, 2001) — Attribution
Non-Commercial No Derivatives — licence.
REFERENCES
Creative-Commons (2001). Creative commons.
Gianpiero, C. (2009). Wata.
Gray, D. (2003). Anonymous coursework submission. In
CompSysTech ’03: Proceedings of the 4th interna-
tional conference conference on Computer systems
and technologies, pages 556–561, New York, NY,
USA. ACM.
LaTeX (1985). Latex project site.
Microsoft (1985). Microsoft windows.
Microsystems, S. (1991). The source for java technology.
Sun-Microsystem, M. (1996). Mysql.
WATA - A System for Written Authenticated though Anonymous Exams
137