A simultaneous spam and phishing attack detection framework for short message service based on text mining approach

Mohd Foozy, Cik Feresa (2017) A simultaneous spam and phishing attack detection framework for short message service based on text mining approach. Doctoral thesis, Universiti Teknikal Malaysia Melaka.

[img] Text (24 Pages)
A Simultaneous Spam And Phishing Attack Detection Framework For Short Message Service Based On Text Mining Approach.pdf - Submitted Version

Download (879kB)
[img] Text (Full Text)
A simultaneous spam and phishing attack detection framework for short message service based on text mining approach.pdf - Submitted Version
Restricted to Registered users only

Download (2MB)

Abstract

Short Messaging Service (SMS) is one type of many communication mediums that are used by scammers to send persuasive messages that will attract unwary recipients. In Malaysia, most sectors such as telecommunication, banking, government, healthcare, and private have taken the initiative to educate their clients about SMS scams. Unfortunately, many people still fall victim. Within the field of SMS detection, only the framework for a single attack detection for Spam has been studied. Phishing has never been studied. Existing detection frameworks are not suited to detect SMS Phishing because these attacks have their own specific behaviour and characteristic words. This gives rise to the need of producing a framework that is able to detect both attacks at the same time. This thesis addresses SMS Spam and Phishing attack detection framework development. 3 modules can be found in this framework, of which are Data Collection, Attack Profiling and Text Mining respectively. For Module 1, the data sets used in this research are from the UCI Machine Learning Repository, the Dublin Institute of Technology (DIT), British English SMS and Malay SMS. The Phishing Rule-Based algorithm is used to extract SMS Phishing. For Module 2, the SMS Attack Profiling algorithm is used in order to produce SMS Spam and Phishing words. The Text Mining module consists of several phases such as Tokenization, Lemmatization, Feature Selection and Classifier. These phases are done with the use of Rapidminer and the Weka data mining tool. Three (3) types of features are used in this framework, which are the Generic Features, Payload Features and Hybrid Features. All of these features are examined and the resulting performance metric used to compare the results is the rate of True Positive (TP) and Accuracy (A). There are four (4) set of results that were successfully obtained from this research. The first result shows that the extraction of SMS Phishing from the SMS Spam class contributes to four (4) enhanced datasets of the UCI Machine Learning Repository, the Dublin Institute of Technology (DIT), British English SMS and Malay SMS. The second results are the SMS Spam and Phishing attack profiling from the enhance UCI Machine Learning Repository, the Dublin Institute of Technology (DIT), British English SMS and Malay SMS. The third and fourth results are obtained from Feature Selection and Classifier phase where Eighty (80) experiments were done to examine the Generic Feature, Payload Features and Hybrid Features. There are five (5) Classification techniques used such as Naive Bayes, K-NN, Decision Tree, Random Tree and Decision Stump. The result of Hybrid Feature accuracy using Rapidminer and Naive Bayes technique is 77.47%, for K-NN: 78.56%, Decision Tree: 57.16%, Random Tree: 57.24% and Decision Stump: 57.16%. Meanwhile, by using Weka the Naive Bayes accuracy rate get 71.45%, K-NN: 81.64%, Decision Tree: 57.10%, Random Tree: 70.64% and Decision Stump: 60.19%. The experiments done using Rapidminer and Weka data mining tool because this is the first survey to detect SMS Spam and Phishing attack at the same time and the results are acceptable. Additionally, the proposed framework also can detect the attack simultaneously using text mining approaches.

Item Type: Thesis (Doctoral)
Uncontrolled Keywords: Data mining, Database searching, Data mining, Statistical methods
Subjects: Q Science > Q Science (General)
Q Science > QA Mathematics
Divisions: Library > Tesis > FTMK
Depositing User: Mohd Hannif Jamaludin
Date Deposited: 26 Mar 2018 08:08
Last Modified: 20 Apr 2022 12:03
URI: http://eprints.utem.edu.my/id/eprint/20626
Statistic Details: View Download Statistic

Actions (login required)

View Item View Item