A Research Of Phishing Detection Technology Based On Deep Learning

In the era of Big Data,network security is still a pivotal topic.there are many illegals take use of the massive information in the network cheating user's trust and making profits,such as the phishing site."phishing" website is similar with real website in URL,web content and layout,users who has no safety consciousness of internet are easily to deceived,and this causing serious consequences.Effectively curb the "phishing site" is a guarantee of network security.Nowadays both at home and abroad have contributed to the study of defensing phishing sites,but there exist many flaws.The existing typical methods of detecting phishing sites,such as black and white list mechanism,detection based on text feature or matching with web image features,classification detection which based on machine learning.However,the blacklist detection is less time-sensitive,the scope of the white list is also inadequate,detection based on the characteristics of fishing site is not satisfactory with accuracy and robustness.In recent years,machine learning has been applied in various fields and achieve some success,especially the application of depth learning in detection and identification can be effectively in improving robustness.In view of the above,In this thesis,we study existing technical methods,and propose a robust phishing site detection method based on deep learning.The main research content of phishing site detection based on the depth of learning are as follows.The extraction of phishing feature is a basis method and also is a key step in the identification of phishing sites,a good feature extraction method plays a crucial role in the subsequent processing of detection.After the investigation of the characteristics of phishing sites and the summary of previous studies,this thesis combines the extracted web pages and web sites to extract the key features of link anomalies and web content anomalies.Firstly we use the URL filter to improve the detection speed and reduce the error rate,secondly we use the similarity judgment which can reduces the detection burden to filtered URL,lastly we preprocesses the URL feature as well as the web page feature and saves the feature vector to detect in the next module.In recent years,deep learning technology has been proposed.It has excellent feature learning ability which makes its application in various fields successfully.Therefore,this thesis studies the classification method of phishing site based on deep learning,and proposes DBN-KNN algorithm model with multi-layer structure,which is applied to the identification of phishing site,and use it to learning,training and classifying the extracted eigenvectors,Finally,according to the classification results to identify the phishing site.In summary,in view of the shortcomings of the existing detection methods,this thesis studies the phishing site detection method based on the deep study.Firstly,crawling the phishing site data,filtering and carrying out similarity detecting URL.Secondly,analyzed and manually extracted the key features of the phishing site with pretreatment.Finally,The depth learning model DBN-KNN is proposed to classify the characteristic vectors and identify the phishing sites.