Design And Implementation Of Malicious URL Real-time Detection System For Mobile Terminal Based On Cloud Computing

The growing popularity of mobile terminals has brought great convenience to people's lives,but also introduced lots of new security problems.More and more attackers shift their attentions to Android-based mobile terminals due to the widespread use of Android system.Malicious URL is one of the most common ways that attackers initiate their malicious activities to Android mobile terminals.By tricking users to click malicious URLs,attackers can automatically download and install malwares on users' terminals to achieve various malicious activities without users' knowledge.Malicious URLs may also link to some phishing pages where attackers pose as legitimate websites to steal users' sensitive.Most of existing security detection methods require the installation of anti-virus programs or intrusion detection systems on mobile terminals.However,most of these methods are constrained by the limitation of memory,storage,computational resources and battery power of mobile terminals.In addition,the limited resources of mobile terminals also limit the real-time demand of such detection methods.Unfortunlately,real-time detection of malicious URLs is a vital factor to protect the interests of users against infringement.The system presents 17 statistical features,and the decision-making forest algorithm is improved.Based on this,a cloud-based real-time detection system for mobile terminal based on cloud computing is designed and implemented.This system mainly aims at the Android mobile terminals.First of all,we propose a set of statistic features to distinguish malicious URLs from legitimate URLs solely based on their own information,include 5 URL inspection features and 12 URL network features.Then,we perform the real-time detection of malicious URL by means of Mahout machine learning platform with three machine learning algorithms:Naive Bayes,Decision Forest and Logistic Regression.This detection system is deployed in a Hadoop cloud.It adopts Hadoop distributed file system(HDFS)to store relevant data,and adopts MapReduce to deal with computing.In this way,it solves the contradiction between resource limitation of mobile terminal and high resource consumption of detection system.We collect various URLs of real mobile Internet by an improved Web crawler,and determine the label of each URL category through a third-party detection system.In this way,we obtain a large-scale dataset of URLs.A series of experiments on this dataset are conducted to evaluate our system.The experimental results show that the system is not only able to detect malicious URLs on Android mobile terminal accurately,but also has a fast detection speed.Thus,our system both satisfies the detection requirement of accuracy and real-time.