Research On Detection Methods Of Financial Phishing Webpage For Mobile Communication

Mobile Internet of China has a huge user base,supporting the rapid development of the mobile phone shopping,mobile Internet banking,mobile payment and other business.Massive users information through the mobile communication network uploaded to a variety of application service providers,users face a serious threat of phishing attacks.At the same time,the number of financial phishing websites has been at the forefront of the number of phishing sites,and phishing has a centralized tendency to focus on the attack,and pseudo base station phishing of financial field has become one of the most serious areas which users suffer losses.In order to effectively deal with the financial phishing attack,targeted design phishing detection program,which is important to timely and effectively identify financial phishing web pages.According to the shortcomings of the existing detection methods,and commonly evading detection means of financial phishing,this paper puts forward and combines the URL features,page content features and logo image characteristics,to establish a set of complete phishing recognition system.The specific work is mainly manifested in the following three aspects:(1)The simulated browser automatically retrieves dynamic web pages,and extracts multiple web content features.Using PhantomJS tools instead of real browser,combined with web automation testing tool Selenium,to achieve dynamic web page crawling and parsing.Getting the text information in the HTML specific tags of the webpage,to match the sensitive text with the multiple pattern matching algorithm AC-SC(AC Suitable for Chinese),and to get sensitive text feature;extracting web page form feature,to enhance the awareness of camouflage means,such as images instead of text.(2)Intercepting the logo image of webpage automatically,to extract image feature.The image is extracted by the PCA-SIFT algorithm with the smallest size,the algorithm has better robustness to logo image stretching,background change and so on.Simultaneously,the algorithm can reduce the dimension of the image feature descriptor,so as to improve the efficiency of image matching.(3)Establishing a complete set of financial phishing web detection system.The URL is required to be filtered by the domain name black and white list before the web feature vector is extracted.The third-party data interface is used to obtain the URL feature,and the 10-dimensional eigenvector is formed with the page feature and logo image feature.The SVM classifier is trained by the feature vector classification rules and to achieve the identification of phishing web pages.The experimental results show that compared with other SVM-based phishing web page detection systems,the proposed method has good relevance and effectiveness in feature selection.At the same time,it reaches 97.43% detection accuracy and obtains false positive rate of no more than 0.86%.