Research Of SMS Botnet Detection Based On Multi-features

The security of mobile internet is facing great challenges as the development of mobile internet and the popular of smart mobile phone.The Short Message Service has the most extensive application for smartphones.This service has also attracted the attention of attackers.Attackers are using SMS to launch a variety of malicious behavior,and one of the most serious hazards is mobile botnets built on SMS.SMS botnet,based on traditional botnet,is specially designed for mobile internet and smart phones.From the huge spreading range and destructive force of traditional botnets,the rapid development of SMS botnet brings a great threat to mobile internet.Therefore,how to effectively detect SMS botnets has become an important topic in botnets.However,at present,researchers have little study on SMS mobile botnets,and the study is mainly focused on the construction of SMS mobile botnets.There are few studies on the detection of SMS botnet,and the traditional botnet detection method is not satisfied with the detection of SMS mobile Botnet.This paper is focused on the construction of SMS detection model.Firstly,this paper analyse the structure and characteristics of SMS botnets to find features.By studying each feature in detail,we evaluate the difficulty and detection ability of these features.We find two types of features,which have good detection value and is easy to draw.These two types of features can be extracted from a group of SMS.These two features are behavior features and textual features.Behavior habits contain two types of features,the time interval and the text size of short messages.Textual features include custom features and feature words.The custom features include embedded telephone numbers,embedded links,the number of special symbols,and the number of spaces.Secondly,we propose a multi-feature SMS botnet detection model based on these two features.The detection model design to determine whether the sender of a short message group is a SMS botnet.Text messages sent by SMS botnet are recorded as bot short message and sent by ordinary users are recorded as ordinary short message.The detection model for SMS botnet detection is transformed into the classification of short message groups.This detection model is mainly divided into two core detection modules.The first detection module uses the entropy of behavioral characteristics to classify short message groups.The second detection modules classify text by text classification.The second detection modules use the naive Bayes algorithm.The second module classify of short message groups that cannot determine the source in the first class detection module.The two detection modules work together to detect whether the sending source of the tested SMS group is SMS botnet.Finally,we design several groups of contrast experiments to test the detection results of the model.The experimental results show that the proposed detection method has a higher accuracy than the pure text feature detection method.