A content-based spam filtering approach using artificial neural networks
Other Title(s)
تصفية البريد المزعج باستخدام الشبكات العصبية الاصطناعية
Dissertant
al-Madhkur, Nur al-Huda Jasim Muhammad Abbas
Thesis advisor
University
University of Baghdad
Faculty
College of Science
Department
Department of Computer Science
University Country
Iraq
Degree
Master
Degree Date
2013
English Abstract
As the popularity of the Internet increased, electronic mails (emails) have become a very common and convenient medium for daily communications.
The spam defined as unsolicited commercial bulk emails, or uninteresting emails has threatened the internet security and email services.
Since the spammers constantly improve their techniques to compromise the spam filters, building a spam filter that can be incrementally learned and adapted became an active research field.
This thesis proposes a spam filtering approach using two “Artificial Neural Networks” ANNs algorithms “Back Propagation” BP and “Optical Back Propagation OBP to identify whether a message is spam or legitimate email based on the content of the message.
These two neural networks should be trained with group of trained samples to distinguish whether a message is spam or legitimate email.
These samples are drawn from Spam-based dataset.
The samples of this dataset should be preprocessed to be in a suitable form that could be understood by neural networks.
The preprocessing operations are features extraction using “Principle Component Analysis” PCA and normalization.
Several experiments are conducted to show the effectiveness of the proposed spam filtering approach and a comparison is made among these experiments with different evaluation measurements.
The results of the tested spam-based dataset feature set size of 100% show that BP and OBP are comparable in terms of accuracy (100%), recall (100%), precision (100%), false positive (0%), and false negative (0%) but OBP takes the least execution time (1.222 seconds).
Also the results show that OBP with 75%, 50%, and 25% feature set size are better than the corresponding BPs in all evaluation measurements
Main Subjects
Information Technology and Computer Science
Topics
No. of Pages
98
Table of Contents
Table of contents.
Abstract.
Abstract in Arabic.
Chapter One : Overview.
Chapter Two : Spam and neural networks.
Chapter Three : Neural networks for spam filtering.
Chapter Four : Experiments and results evaluation.
Chapter Five : Conclusions and future work.
References.
American Psychological Association (APA)
al-Madhkur, Nur al-Huda Jasim Muhammad Abbas. (2013). A content-based spam filtering approach using artificial neural networks. (Master's theses Theses and Dissertations Master). University of Baghdad, Iraq
https://search.emarefa.net/detail/BIM-605837
Modern Language Association (MLA)
al-Madhkur, Nur al-Huda Jasim Muhammad Abbas. A content-based spam filtering approach using artificial neural networks. (Master's theses Theses and Dissertations Master). University of Baghdad. (2013).
https://search.emarefa.net/detail/BIM-605837
American Medical Association (AMA)
al-Madhkur, Nur al-Huda Jasim Muhammad Abbas. (2013). A content-based spam filtering approach using artificial neural networks. (Master's theses Theses and Dissertations Master). University of Baghdad, Iraq
https://search.emarefa.net/detail/BIM-605837
Language
English
Data Type
Arab Theses
Record ID
BIM-605837