‘CodeAliker’ - Plagiarism Detection on the Cloud

Advanced Computing: An International Journal ( ACIJ ), Vol.3, No.4, July 2012
DOI : 10.5121/acij.2012.3403 21
‘CodeAliker’ - Plagiarism Detection on the Cloud
Nitish Upreti1
and Rishi Kumar2
Department of Computer Science and Engineering, AMITY University, Noida, India.
nitishupreti@gmail.com
Department of Computer Science and Engineering, AMITY University, Noida, India.
rishikumar182000@gmail.com
Abstract
Plagiarism is a burning problem that academics have been facing in all of the varied levels of the
educational system. With the advent of digital content, the challenge to ensure the integrity of academic
work has been amplified. This paper discusses on defining a precise definition of plagiarized computer
code, various solutions available for detecting plagiarism and building a cloud platform for plagiarism
disclosure.
‘CodeAliker’, our application thus developed automates the submission of assignments and the review
process associated for essay text as well as computer code. It has been made available under the GNU’s
General Public License as a Free and Open Source Software.
Keywords
Plagiarism, String matching, Cloud Computing
1.Introduction
An insightful look into the scenario of academic integrity and its implications give us the major
motivation for pursuing the subject. The issue holds utmost significance as the intellectual
standards of an individual pursuing an academia a reestablished around his ability to produce
authoritative work. Plagiarism is thus lethal. Every year a large number of students and scholars
submit a huge volume of material to their respective mentors and professors. Due to the sheer
amount of text involved, a manual scrutiny is infeasible. Analyzing the situation, we found no
existing work in the public domain that solved the problem faced by educational institutes
worldwide. Most of the alternatives were either closed source or catered to only a fraction of the
entire problem. Working on this issue, at the outset we explore the sensitive aspect of
classification of documents as ‘authentic’ or ‘plagiarized’. We then analyze numerous
approaches to Plagiarism detection. Advancing then to our chief goal of implementing an
engine and leveraging the cloud platform for scalable and robust plagiarism detection. Alex
Aliken’s MOSS[1] is chosen as the key approach for building the application. Result and
conclusion follow where we present our observations and learning.
2. Classification of Text
Broadly categorizing, the nature of text submitted to such a system can either be an Essay that is
plain text in language or computer code in any of the popular language such as C, C++, Java or
Ruby for instance. It is easy to figure out whether an essay text has been plagiarized however
source code copying is a delicate issue with mostly a fine line drawn between ‘code reuse’,

Advanced Computing: An International Journal
‘collaboration’, ‘non-citation’ and ‘plagiarism’
an OOP system and ‘Don’t re-invent the wheel
more blurred. With hardly any definition
identifying copying instances is infeasible. Hence
little work has been done on the topic; the only concrete input comes from the work of
Cosma and Mike Joy [2]. Their work follows a
finding the right answers and opinions
to have our own precise judgment on
assignment submitted consists of
in almost all the sophisticated code bases, could be a major potential resource for
plagiarism instances. Nonetheless
looking false positives as Copyright statements occur
of CodeAliker we choose to strip off comments
we address a multitude of other questions. Is using an external library or API an instance of
plagiarism? For most of the cases
are a central part of any sophisticated piece
import statements and library includes.
submission; code for the design is
manually is much more effective than
3. Approaches to Plagiarism Detection
Various different approaches to Plagiarism detection exist and their performance and speed
to a great extent. Also certain plagiarism detection sche
specific structure and nature. A rich t
Plagiarism
Web Scrapping Based
String Matching
citation’ and ‘plagiarism’. With learning themes such as ‘Code Reuse’ in
invent the wheel’ code philosophies, the distinction are
With hardly any definition in place designing a system capable of accurately
nstances is infeasible. Hence concrete definitions need to be in place.
topic; the only concrete input comes from the work of
[2]. Their work follows a survey-based approach in the U.K academics
finding the right answers and opinions. However for implementing a practical solution we need
judgment on the problem rather than a crude hypothesis
submitted consists of comments and the actual source code. Comments, which occur
in almost all the sophisticated code bases, could be a major potential resource for
onetheless they present a major pitfall and could lead on to suspicious
Copyright statements occur frequently as comments. For the purpose
we choose to strip off comments so as to avoid any such issues. Moving forward
t of the cases we found that library use without citation is legitimate
sophisticated piece of computer program. CodeAliker thus filters out
import statements and library includes. An intuitive User Interface design can also be a part of
ode for the design is put under scrutiny by CodeAliker but looking at the design
manually is much more effective than plain UI code checking.
Approaches to Plagiarism Detection
Various different approaches to Plagiarism detection exist and their performance and speed
Also certain plagiarism detection schemes are more suitable for data
A rich taxonomy can be summarized in the diagram below
Figure 3.1
Plagiarism Detection Approaches
Local Database Based
Non Sturctured
Fingerprinting
Techniques
Matching
Parameterized
Techniques
Structured
ACIJ ), Vol.3, No.4, July 2012
22
With learning themes such as ‘Code Reuse’ in
distinction are even
m capable of accurately
need to be in place. A
topic; the only concrete input comes from the work of Georgina
in the U.K academics for
However for implementing a practical solution we need
rather than a crude hypothesis. Code
Comments, which occur
in almost all the sophisticated code bases, could be a major potential resource for identifying
and could lead on to suspicious
For the purpose
Moving forward,
legitimate as they
CodeAliker thus filters out
an also be a part of
but looking at the design
Various different approaches to Plagiarism detection exist and their performance and speed vary
data set with a
axonomy can be summarized in the diagram below.
Based
Sturctured

23
Web Scrapping based approaches use the World Wide Web to check for Plagiarism instances
from a large corpus of data. The scope of Web Scrapping is huge and lots of published work
exists on such systems. Our focus for this research is on systems based on a local database
compiled from assignments submitted by students taking the classes and past year submissions.
Local Database Based Approaches can be either Structured or Non Structured. The Structured
approach creates a graph model of information in the document. This approach is used mostly
with code-based assignments.
Non-Structured techniques are the most popular ones and are useful on a wide variety of text
material. They are classified based on the algorithm used. Document Fingerprinting, String
Matching and Parameterized Matching are the popular ones [3].
Tools based on the fingerprint approach work by creating “fingerprints” for each file which
consist statistical information about the file, such as average number of terms per line, number
of unique terms, and number of keywords [4].The DUP tool [5] is based on a parameterized
matching algorithm, which detects identical and near-duplicate sections of source-code, by
matching source-code sections whose identifiers have been substituted (renamed) systematically
[3].
String Matching algorithms are quite popular and effective. MOSS [1], (YAP3) [6], JPlag [7],
and Sherlock [8] are some of the popular ones available. CodeAliker is based on MOSS[1] that
employs string-matching algorithms using k-grams, where a k-gram is an adjacent substring of
length k. Winnowing, a local fingerprinting algorithm is also used to ensure matches of certain
length are detected.
4. Designing the Engine with Ruby
There were various motives for choosing MOSS as the core for CodeAliker’s engine. Also
Ruby was used to implement the engine after considering several important factors. The
language provides excellent text processing libraries, encourages an agile development
methodology and Test Driven Development (TDD). Moreover it is ready for the web with
excellent frameworks available.
MOSS is highly effective for plagiarism detection with text of different nature. It can also be
scaled to handle a large volume of data. MOSS also guarantees matches of certain length to be
detected [1].
The engine consists of three major modules: Text Filter, Hasher and Winnower. All of the
components can be customized with easy to write configuration files.
The text filter has a key role to play when processing code assignments. Based on the approach
MOSS suggests, the comments are stripped off, text is lowercased, identifiers are replaced with
a dummy symbol, language specific keywords are removed and punctuations with no semantic
meanings are stripped off. Filtered text with noise eliminated is thus obtained.
The filtered text is then fed to a Hasher that calculates hashes for the given text. A rolling hash
function based on the famous Rabin Karp Algorithm is employed to calculate hashes quickly.
With each hash value calculated, the corresponding line number where the text occurred is
stored. This aids later in presenting user with the information regarding the instances where
plagiarized text is present.
The Winnower is an implementation of the ‘Robust Winnowing’ algorithm defined by MOSS.A
set of hash is chosen to be as the finger print of a document. Line number information is still
preserved.

24
Winnower needs to be configured with parameters value ‘k’ for k-gram, a threshold value ‘t’
and a modulus value ‘q’. If there is a substring match at least as long as the guarantee
threshold, ‘t’, then this match is detected, and we do not detect any matches shorter than
the noise threshold, ‘k’ [1].The hash values computed are two large and hinder a
scalable implementation; hence a value ‘q’ is used as the modulus.
For CodeAliker we found the sweet spot with the values 5(k), 8(t) and 10001(q) respectively.
The documents are compared based on the final fingerprints, with plagiarism instance being
reported line by line. Check for essay based assignment is surprisingly similar with the Filter
step being omitted.
5. Building a Cloud Application
The most interesting part of our research is to build a cloud application for the engine. For
building the web application we employ the Ruby on Rails platform.
Ruby on Rails, a full stack framework for Ruby is excellent for agile development and
sustainable productivity. It boasts a high modular design, excellent package management
capabilities, database abstraction with ORM(Object Relational Mapping) library and ease of
deployment.
The application is built with the MVC (Model – View – Controller) design pattern inherent on
the Rails Framework. CodeAliker aims to ease the workflow involved by automating the entire
process. To achieve this, an authentication-based system is introduced for the professors where
assignments for each class they take are available to them as a separate bunch.
The professor can mark any on the assignment as primary and check with respect to that
assignment all the possible plagiarized instances. Thus getting a complete view of the scenario
effortlessly in a non ad-hoc fashion. Academics can also manually supervise the submission and
reviews.
While the traditional delivery of software services have been mainstream, bringing the cloud
into perspective changes the entire scenario. Cloud tends to centralize our resources, code base
and data onto an always-available depot. Hardware resources can be accurately utilized, load on
a high demand system can be catered to and system can be easily scaled. Configuration
management is also made effortless. Any organization with requirements for sucha system is in
need of the cloud. The platform allows changes to be pushed onto codebase with a push of a
button, rather than relying on extensive upgrade packs.
For plagiarism detection in a university or an academic institute the needs are critical and point
towards the cloud. The computing requirements are thus addressed. Also our system is
centralized and easily scaled.
For hosting CodeAliker on the cloud, Heroku, a cloud platform has been employed.
The source code is available to public here on: https://github.com/Myth17/CodeAliker
The application is available for free use at: http://codealiker.heroku.com/

Advanced Computing: An International Journal
6. Results
The results for CodeAliker display
suspected plagiarism instances are previewed
clearer picture, the plagiarized instances are marked with the line numbers in order of aid
manual scrutiny and presenting a
7. Conclusion
We have analyzed the entire scenario of Plagiaris
and solutions for developing an
research being our ability to define
different approaches towards plagiarism detection, practical implementation of the MOSS
engine with fine tuned parameters
platform.
Figure 6.1
Figure 6.2
results for CodeAliker display the assignment marked as primary to the left while the
suspected plagiarism instances are previewed stacked onto each other in the right. To present a
iny and presenting a more cohesive report.
lyzed the entire scenario of Plagiarism detection, while figuring out the
olutions for developing an application for the purpose. Major accomplishment of the
our ability to define a precise definition of code plagiarism, understanding
towards plagiarism detection, practical implementation of the MOSS
parameters and building a scalable web application hosted on a cloud
ACIJ ), Vol.3, No.4, July 2012
25
the assignment marked as primary to the left while the
the right. To present a
while figuring out the problems
Major accomplishment of the
understanding
towards plagiarism detection, practical implementation of the MOSS
hosted on a cloud

26
References
[1] Saul Schleimer, Daniel S. Wilkerson and Alex Aiken, “Winnowing: Local Algorithms for Document
Fingerprinting”, SIGMOD '03 Proceedings of the 2003 ACM SIGMOD international conference on
Management of data, pp 76-85, 2003.
[2] Georgina Cosma and Mike Joy, “ Towards a definition of Source Code Plagiarism”, IEEE
TRANSACTIONS ON EDUCATION, VOL. 51, NO. 2, MAY 2008.
[3] Georgina Cosma and Mike Joy, “An Approach to Source-Code Plagiarism Detection and
Investigation Using Latent Semantic Analysis”, IEEE TRANSACTIONS ON COMPUTERS, VOL.
61, NO. 3, MARCH 2012 379.
[4] M. Mozgovoy, “Desktop Tools for Offline Plagiarism Detection in Computer Programs,”
Informatics in Education, vol. 5, no. 1, pp. 97- 112, 2006.
[5] B. Baker, “On Finding Duplication and Near-Duplication in LargeSoftware Systems,” Proc. IEEE
Second Working Conf. Reverse Eng.,pp. 85-95, 1995.
[6] M.J. Wise, “YAP3: Improved Detection of Similarities in Computer Program and Other Texts,”
Proc. 27th SIGCSE Technical Symp.,pp. 130-134, 1996.
[7] L. Prechelt, G. Malpohl, and M. Philippsen, “Finding PlagiarismsAmong a Set of Programs with
JPlag,” J. Universal ComputerScience, vol. 8, no. 11, pp. 1016-1038, 2002.
[8] M. Joy and M. Luck, “Plagiarism in Programming Assignments,” IEEE Trans. Education, vol. 42,
no. 2, pp. 129-133, May 1999.
Authors
Nitish Upreti is computer science student at AMITY School of Engineering and
Technology, Noida, India. His fields of interest include Algorithm Design, Artificial
Intelligence and building scalable web applications.
Rishi Kumar is computer science faculty at Amity School of Engineering and
Technology, Noida, India. His fields of interest include Artificial Intelligence,
Expert System & Image Processing.

‘CodeAliker’ - Plagiarism Detection on the Cloud

More Related Content

What's hot (20)

Similar to ‘CodeAliker’ - Plagiarism Detection on the Cloud (20)

More from acijjournal (20)

Recently uploaded (20)

‘CodeAliker’ - Plagiarism Detection on the Cloud