Detecting Phishing Web-pages Based On The Spatial Database And Visual Layout Features

Phishing attack is becoming a critical important issue in our daily life, which belongs to information security area. Criminals are trying to steal victims'accounts, passwords and other personal information. These kinds of fraud sites do great harm to target users all around the world, which is becoming a serious threat to the well-development of the Internet world. Thus it is necessary to do research on phishing detection and develop software tools with high accuracy and efficiency. The main work and contribution come out of the thesis are:Firstly, from the sites'topology perspective: the topology of a phishing site is very simple, usually contains several web pages, while that of the legal website is so complicated that usually has more than hundreds and thousands of internal links. So the thesis put forth a phishing website detection method based on topology of website. A crawler is used to fetch some web pages according to the internal links in the site, and thus the topological features can be extracted. Then a classifier is trained on the extracted topological features to get the final results. The experiments show that the topology-based method can detect phishing websites effectively with high accuracy and high recall rate.Secondly, from the visual similarity perspective: a suspicious web page often looks like a legal normal page. So the thesis proposes a new anti-phishing method which is based on spatial database and visual similarity. The spatial layout features of the legal web page are extracted, which are then saved in the spatial database in order to be indexed uniformly. Then the similarity detection of the webpage is taken based on the spatial database. The experiments show that this method is capable of processing large datasets with high accuracy, and can reduce the time-consuming significantly.