Research On Phishing Websites Detection Based On Optimal Feature Selection And Neural Network

With the development of the Internet and the frequent exchange of data,the security between information exchange becomes more and more important.The phishing attack has become the fastest growing attack mode because of its short survival and great harm.Phishing attacks are an attack that uses social engineering and technical spoofing to obtain user identity data and financial account data.The most common way is to send fake website links to users on the network and entice users to click,to monitor and intercept the user's private information without authorization,which brings huge economic losses to users.Therefore,establishing a mechanism for quickly detecting and processing phishing websites can effectively stop the harm caused by phishing attacks in a timely manner.Since the traditional phishing website detection technology lacks the active learning ability of large-scale data sets,the autonomous extraction feature of machine learning algorithms has become the mainstream detection technology.The key to this detection method is the construction of features and the choice of classification algorithms.This thesis conducts an in-depth study on the detection of relevant features of phishing websites.Due to the variety of features of phishing websites,the characteristics of manual extraction often rely on empirical knowledge.This may lead to some features that cannot be used to effectively distinguish phishing websites,and also bring inefficiencies in detection.However,these useless features also affect the training effect of the machine learning model.As a result,the trained model cannot accurately predict and detect the phishing website.The machine learning algorithm model also shows different effects in detecting phishing websites.This thesis compares the classification effects of commonly used machine learning models with experiments and selects an efficient neural network model as the algorithm model of the detection framework.Based on the above analysis,based on the optimal feature selection method,an effective neural network detection model OFS-NN(Optimal Feature Selection-Neural Network)is proposed to detect phishing websites.The main work of this thesis is as follows:(1)Based on the analysis of the existing phishing technology principle and the existing phishing website detection model advantages and existing defects,and by comparing various machine learning detection models,the neural network classification model suitable for phishing websites is obtained.The neural network model has high precision,strong robustness and strong fault tolerance for noise data.In addition,the neural network model can simulate complex nonlinear relationships and better learning ability,and can predict unknown types of phishing websites.(2)This thesis extracts the corresponding sensitive features by extracting the URL information,HTML information and DNS information of the website.However,the useless features will affect the detection effect and efficiency of the model.This thesis proposes an optimal feature selection index FVV(Feature Validity Value)to eliminate the useless features.Based on the calculation of the effective value of each feature,a threshold is set to eliminate some useless features to select the optimal feature set suitable for training the machine learning algorithm.Compared with the Gain(information gain)index,the proposed index has better feature selection ability.This thesis gives the optimal feature selection algorithm based on the FVV index,which improves the performance of the model training process and the detection process.(3)Combining the optimal feature extraction algorithm and neural network algorithm,this thesis proposes a neural network phishing detection model OFS-NN based on optimal features.In this thesis,by selecting the optimal feature set,the optimal neural network classifier is constructed to classify and predict the phishing website.And the black and white list mechanism is introduced when detecting the phishing website to improve the detection efficiency.The experimental results show that the proposed OFS-NN model provides an effective solution for the prediction and detection of phishing websites.The model has high precision and powerful generalization ability,which can effectively identify multiple types of phishing websites.