Research And Application Of BP Neural Network Based On Improved Grey Wolf Algorithm

The fast-developing information age has brought great convenience to people's lives.More and more people are used to doing things online,such as online shopping and online payment.This phenomenon has become a normal state.It is more convenient to obtain all kinds of information through the network,which provides a fertile living ground for phishing.The ubiquity of phishing has brought great threats to people's production and life.The property security of netizens has been violated,which is not conducive to economic development and social stability.These drawbacks have made the study of phishing detection technology a key research topic.At present,the technology of phishing detection has the disadvantages of low efficiency,poor stability and poor generalization.Therefore,it has become particularly important for the research and improvement of phishing detection technology.Based on the research of machine learning algorithms,this thesis optimizes BP neural network by improved grey wolf algorithm and applies it to the detection of phishing websites,which has achieved good results.The main work and contributions of this thesis are as follows:(1)This thesis summarizes and analyzes the attack methods of phishing websites and existing related detection technologies.A machine learning algorithm was chosen as the classification algorithm for phishing websites.At the same time,by evaluating the advantages and disadvantages of various machine learning classification algorithms,BP neural network algorithm is selected as the predictive classifier of phishing websites.Based on the BP neural network algorithm,it is easy to fall into the local minimum and the slow convergence speed.This thesis combines the gray wolf algorithm with strong search ability to improve the algorithm.(2)For the gray wolf algorithm,this thesis first analyzes the principle and shortcomings of the gray wolf algorithm.This thesis also improves the local optimal defects of the grey wolf algorithm by optimizing the control factor and adding variable proportional weight.Then,performance testing and result analysis were performed on the improved gray wolf algorithm(IMGWO).Experiments show that the improved gray wolf algorithm(IMGWO)has improved convergence accuracy and stability.(3)The improved gray wolf algorithm(IMGWO)is combined with the BP neural network algorithm.The optimal location of the grey wolf algorithm is used as tihe weight and threshold selection of the BP neural network to form a new IMGWO-BP classification model.Experiment and test the IMGWO-BP classification model.Experiments show that the IMGWO-BP classification model improves classification accuracy and efficiency.(4)The IMGWO-BP classification model is applied to the actual detection of phishing websites.Firstly,the related process of phishing website detection is designed and analyzed,including pre-judging of URL and feature extraction.Secondly,optimization and analysis of proposed model,including parameter setting and double feature evaluation method.To further improve the recognition rate of phishing website URLs.Finally,the proposed model is tested and analyzed.It mainly includes the selection of test dataset,the test of IMGWO-BP model recognition effect,the test of the recognition effect of DIGWO-BP model after adding dual feature evaluation,and the comparison of DIGWO-BP model with other models.The experimental results show that the DIGWO-BP classification model proposed in this thesis has the advantages of good effect,high efficiency and strong stability in predicting phishing websites.