AN ANALYSIS OF EFFECTIVE ANTI SPAM PROTOCOL USING DECISION TREE CLASSIFIERS

IJSRD - International Journal for Scientific Research & Development| Vol. 2, Issue 09, 2014 | ISSN (online): 2321-0613
All rights reserved by www.ijsrd.com 1
An Analysis of Effective Anti Spam Protocol Using Decision Tree
Classifiers
Sarasu.S 1
Sathees Kumar.B2
1
BCA, MSc., MPhil
2
MCA, MPhil, SET, Assistant Professor
Abstract— As the internet usage increases in day to day
activities, there is an inherent corresponding increase in
usage of communication through it with email being the
mainstay or rather in the forefront of modern day
communication methodologies for businesses and general
persons as well. This has led to get customer attention in the
form of unwanted and unsolicited bombarding of the
customers mail accounts with advertisements, offers,
phishing activities, viruses, worms, trojans, generating hate
crimes, making the customer to part with sensitive
information like passwords, and other media as well which
is known as spam. Spam is mass mailing or flooding of
mail account servers with unwanted trash data causing
damage some times. Spam filters have been in use from the
time such mail flooding happens. Most of the spam filters
are manual meaning which the user after identifying a mail
in his account blocks the sender and henceforth the system
will not allow mails to the inbox from such addresses.
However the spammers are resilient and send spam mails
from different identities and flood the inboxes. This study
focuses on algorithms and data mining techniques used to
unearth spam mails. They filter the inbox mails as they
arrive at the server depending on certain rules which are
already defined known as supervised learning methods.
Such technologies are known as knowledge engineering
techniques. Here a decision classifier is used to train such
mails with varying words to filter and identify the words in
the mail as spam. The Decision Tree model is used to
analyze the mails and identify spam mails and block them.
The number of mails sent, content, subject, type whether
reply or forward, language etc. are identified using the
decision classifier like Naves Bayes and analyzed
accordingly to filter the emails.
Keywords: Spam, Phishing, Email, Clustering, Decision
Tree
I. INTRODUCTION
Spam is unsolicited commercial email sent in bulk; it is
considered an intrusive transmission. These bulk messages
often advertise commercial products, but sometimes contain
fraudulent offers and incentives. Due to the nature of
Internet mail, spammers can flood the net with thousands or
even millions of unwanted messages at negligible cost to
themselves; the actual cost is distributed among the
maintainers and users of the net. Their methods are
sometimes devious and unlawful and are designed to
transmit the maximum number of messages at the least
possible cost to them. Unfortunately, these emails impose a
significant burden upon recipients. Due to the dramatic
increase in the volume of spam over the past year, many
email users are searching for solutions to this growing
problem. The huge amount of unwanted email has led to
significant decreases in worker productivity, network
throughput, data storage space, and mail server efficiency.
In large organizations, a considerable portion of the time of
each worker is spent reviewing and deleting the spam itself,
leading to a decrease in productivity. The increased network
traffic has a deleterious effect on network performance, in
general, and on the organization’s mail server(s), in
particular. Also, data storage space is consumed by the need
to store the large volume of mail.
II. PURPOSE OF SPAM
The motivation behind spam is to have information
delivered to the recipient that contains a payload such as
advertising for a (likely worthless, illegal, or non-existent)
product, bait for a fraud scheme, promotion of a cause, or
computer malware designed to hijack the recipient’s
computer. Because it is so cheap to send information, only a
very small fraction of targeted recipients — perhaps one in
ten thousand or fewer — need to receive and respond to the
payload for spam to be proﬁtable to its sender ]. A decade
ago, the mechanism, payload, and purpose of spam were
quite transparent. The majority of spam was sent by ―cottage
industry‖ spammers who merely abused social norms to
promote their wares.
III. ANTI-SPAM TECHNIQUES / APPROACHES
Anti-spam methods can be grouped into a few, fairly well
defined, categories, though only some of these methods are
currently in use. There are two aspects to the response to
spam. The most commonly discussed problem relates to the
ability to distinguish between spam and legitimate email.
For a large percentage of email, the decision is easy, which
can easily be identified by more than half of all email as
either definitely legitimate (white) or definitely spam
(black). It is the rest that is the most difficult to handle. We
call these mails as ―gray mail‖.
The second issue for any comprehensive spam
solution is the proper response to black and gray email. For
confirmed spam, the solution is often easy to simply delete
them. However, there may be instances where other or
additional actions are appropriate. Some possibilities are:
(1) Forward the spam to the abuse department at the
domain of the originator.
(2) Reply to the originator voicing displeasure at
receiving spam.
(3) Reply to the originator to advise them that the
email was not delivered.
(4) Report the spam to a spam gathering station.
This list is not exhaustive and multiple responses
may be appropriate in some situations. For gray mail, the
appropriate response is unclear. Thus, the goal of a
comprehensive anti-spam product is to be able to identify
every email as either white or black with a very high
probability of accuracy.

An Analysis of Effective Anti Spam Protocol Using Decision Tree Classifiers
(IJSRD/Vol. 2/Issue 09/2014/001)
IV. IMPLEMENTATION
A. Spam Filter Inputs and Outputs
A spam filter with perfect knowledge might base its decision
on the content of the message, characteristics of the sender
and the target, knowledge as to whether the target or others
consider similar messages to be spam, or the sender to be a
spammer, and so on. But perfect knowledge does not exist
and it is therefore necessary to constrain the filter to use well
defined information sources such as the content of the
message itself, hand-crafted rules either embedded in the
filter or acquired from an external source, or statistical
information derived from feedback to the filter or from
external repositories compiled by third parties.
The desired result from a spam filter is some
indication of whether or not a message is spam. The
simplest result is a binary categorization spam or non-spam
which may be acted upon in various ways by the user or by
the system. We call a filter that returns such a binary
categorization a hard classifier. More commonly, the filter is
required to give some indication of how likely it considers
the message to be spam, either on a continuous scale (e.g., 1
= sure spam; 0= sure non-spam) or on an ordinal categorical
scale (e.g., sure spam, likely spam, unsure, likely non-spam,
sure non-spam). We call such a filter a soft classifier. Many
filters are internally soft classifiers, but compare the soft
classification result to a sensitivity threshold t yielding a
hard classifier. Users may be able to adjust this sensitivity
threshold according to the relative importance they ascribe
to correctly classifying spam vs. correctly classifying non-
spam.
B. Filter
A filter may also be called upon to justify its decision; for
example, by highlighting the features upon which it bases is
classification. The filter may also classify messages into
different genres of spam and good mail. For example, spam
might be advertising, phishing or a Nigerian scam, while
good email might be a personal correspondence, a news
digest or advertising. These genres may be important in
justifying the spam/non-spam classification of a message, as
well in assessing its impact.
The typical use of an email spam filter from the
perspective of a single user. Incoming messages are
processed by the filter one at a time and classified as ham (a
widely used colloquial term for non-spam) or spam. Ham is
directed to the user’s inbox which is read regularly. Spam is
directed to a quarantine file which is irregularly (or) never)
read but may be searched in an attempt to find ham
messages which the filter has misclassified. If the user
discovers filter errors either spam in the inbox or ham in the
quarantine — he or she may report these errors to the filter,
particularly if doing so is easy and he or she feels that doing
so will improve filter performance.
V. EVALUATION
The filter is on-line in that it processes one message at a
time, classifying each in turn before examining the next.
Furthermore, it is passive in that it makes use only of
information at hand when the message is examined. Variants
of this deployment are possible, only some of which have
been systematically investigated:
Batch filtering, in which several messages are
presented to the filter at once for classification. This method
of deployment is atypical in that delivery of messages must
necessarily be delayed to form a batch. Nevertheless, it is
conceivable that filters could make use of information
contained in the batch to classify its members more
accurately than on-line.
Batch training, in which messages may be
classified online, but the classifier’s memory is updated only
periodically. Batch training is common for classifiers that

involves much computation, or human intervention, in
harnessing new information about spam.
Just-in-time filtering, in which the classification of
messages is driven by client demand. In this deployment a
filter would defer classification until the client opened his or
her mail client, sorting the messages in real-time into inbox
and quarantine.
The vast breadth of the spam ecosystem and
possible abatement techniques render impossible the direct
measurement of these quantities; there are simply too many
parameters for any single evaluation or experiment to
measure all their effects at once. Instead, we make various
simplifying assumptions which hold many of the parameters
constant, and conduct an experiment to measure a quantity
of interest subject to those assumptions. Such experiments
yield valuable insight, particularly if the assumptions are
reasonable and the quantities measured truly illuminate the
question under investigation. The validity of an experiment
may be considered to have two aspects: internal validity and
external validity or generalizability. Internal validity
concerns the veracity of the experimental results under the
test conditions and stated assumptions; external validity
concerns the generalizability of these results to other
situations where the stated assumptions, or hidden
assumptions, may or may not hold. Establishing internal
validity is largely a matter of good experimental design;
establishing external validity involves analysis and repeated
experiments using different assumptions and designs.
VI. RESULTS AND DISCUSSIONS
The time feature is an important task and the proposed
mechanism processes in fewer time frames as shown in the
table given below. It is less than half which means it is
effective by more than 50% of the earlier solutions which is
a heartening factor.
Factor Existing Proposed
Time 0.56 0.125
A. Table for Time Factor to Detect Spam
Chart Showing Time Factor to Detect Spam
Training time for the spam detector was also taken
into account and this was found to take less time for training
and thus the training time was found to be less for the
proposed mechanism. Hence the detection would start
immediately after the training finishes. The below table and
graph illustrates this.
Training Time 42 20
B. Table for Training Time Factor
Chart Showing Training Time Factor
The next factor considers accuracy of the spam
detection and this was also monitored and mapped into an
appropriate table and graph which again showed that the
proposed hybrid mechanism was resilient and accurate in
terms of accuracy of detecting spam mails.
ACCURACY 65% 94%
Table showing Accuracy of the Spam Detector.
Chart showing Accuracy of the Spam Detector.
Thus the evaluation based on spam for the
proposed model shows the accuracy, training time and
detection time with less filtration techniques and resilient to
spam attacks.
VII. CONCLUSION
In the modern communication world email spam is fast
becoming a serious threat for regular users of emails,
businesses and corporate firms. This study has shown how
to filter the spam that causes trouble by the proposed
suitable hybrid algorithm on the set of emails which consist
of spam and ham mails collected from our sources. Result
analysis on data shows that the proposed method is effective
and also suitable for all mails. It also does not show false
positives or negatives.
This model can be extended in the form of web
services as part of the future enhancements. This will enable
different mail service providers to incorporate the spam
features into their service without actually writing the code
but directly implementing them.
0.56
0.125
0
0.2
0.4
0.6
Existing Proposed
Time
0
50
Existing Proposed
42
20
Training Time in msecs

REFERENCES
[1] P.Sudhakar, G.Poonkuzhali, K.Thaigarajan,
K.Sarukesi, International Journal of Compuers, Issue
3, Volume 5, 2011, P. 332-345
[2] Almeida T, Yamakami A, Almeida J (2009)
Evaluation of approaches for dimensionality
reduction applied with Naive Bayes anti-spam filters.
In: Proceedings of the 8th IEEE international
conference on machine learning and applications,
Miami, FL, USA, pp 517–
[3] Cormack G (2008) Email spam filtering: a systematic
review. Found Trends InfRetr 1(4):335–455
[4] Machine Learning Techniques in Spam Filtering
Konstantin Tretyakov, kt@ut.ee, Institute of
Computer Science, University of Tartu, Data Mining
Problem-oriented Seminar, MTAT.03.177, May
2004, pp. 60-79.
[5] A Study on Email Spam Filtering Techniques,
Christina V et. all. International Journal of Computer
Applications (0975 – 8887) Volume 12– No.1,
December 2010, pp.07-09

AN ANALYSIS OF EFFECTIVE ANTI SPAM PROTOCOL USING DECISION TREE CLASSIFIERS

More Related Content

What's hot (20)

Similar to AN ANALYSIS OF EFFECTIVE ANTI SPAM PROTOCOL USING DECISION TREE CLASSIFIERS (20)

More from ijsrd.com (20)

Recently uploaded (20)

AN ANALYSIS OF EFFECTIVE ANTI SPAM PROTOCOL USING DECISION TREE CLASSIFIERS