Detection of Spam in SMSs based on Naïve Bayes Technique

Nowadays cell phones are easily available for all types of people. One of the main capabilities which is introduced by these devices is their special messaging system which is called Short Message Service (SMS). SMSs make it possible to transfer users’ messages to their audiences with the least possible cost. Although SMS service seems like an appropriate advertising tool, but on the downside the issue of spam messages is born. Spams might even impose unnecessary cost for users. Therefore, proper detection and elimination of such messages is necessary. For elimination of spams, it is just enough to detect them correctly and then prevent from their transmission. In this research a new technique has been proposed for detection of spams based on data mining methods. In order to make spam detection, text processing in RapidMiner software was used and 4-gram features were extracted in this software. In order to have a proper classification to reduce the features in the process, feature selection had been used by correlation method. Finally, Naïve Bayes technique were used for the classification of data. Based on precision, accuracy and recall measure, the acquired results are 96.14, 95.75 and 99.97 respectively.

Keywords:

Spam , SMS , feature selection , Naïve Bayes classifier