Development And Implementation Of The Identifying Network Robot Software Base On Users' Mouse Operation

Nowadays,online shopping has been popularized.E-commerce websites display a lot of information about commodities,such as price,transaction records,etc.These valuable information will attract competitors'online robots to patronize.These web robots can adversely affect website performance and information security.Network robot,also known as web crawler,which is a program that automatically collects website information according to certain rules.It is also divided into the regular and the irregular.The regular web crawler,such as Baidu,Google,are to facilitate the dissemination of information.The irregular will malicious steal sensitive data or increase website burdens.More than 90%of the traffic in the entire network world is from the network robots.In order to protect sensitive data and reduce the burden of the site,a software is needed which can identify whether the requesting site data is real users or network robots.This thesis is a way to identify the network robots by the user's mouse operation mode,and this way will protect the sensitive data of the website.Conventional means are artificial identification,authentication through User-Agent information,analysis of log identification,and analysis of IP request authentication and so on.These methods are simple and effective,but the automated level is not enough,resulting in too much investment in manpower,time and so on,which causes the high cost.We also use the mouse and keyboard behavior to extract relevant eigenvalues,and then use machine algorithm to identify.The effect is not good enough.Because of the limitation of hardware devices,we can not get a better model at the early stage.The way of this thesis is based on the data of the mouse operation generated by the user when browsing the website,and according to the differences of the mouse operation behavior between the actual human and the network robot,the mouse operation feature is abstracted and the model is established,and the feature model of the mouse operation is solidified.And then through the model evaluation to distinguish whether the new mouse operation is a network robot.The implementation of this method is mainly divided into:(1)the collection of mouse operating behavior,and the abstraction of features.(2)Establish and solidify mouse model;(3)Evaluation and judgment of new mouse operation;(4)Software docker package.Acquisition:Through the writing of JavaScript scripts,the ion of collection and features is realized,and the data transmission using ajaxa technology is realized.Establishing a model:By comparing two implementations of Stacking and TensorFlow,and then choose the way of high recognition accuracy.Evaluation:Using the API interface method to evaluate and identify new mouse operations.Encapsulation:Use Docker to package the software,improve the efficiency of development and deployment,and improve performance through load balancing.This thesis finally realizes the authentication network robot software by API interface.Through the high concurrency evaluation of the test data set,the accuracy of the identification results can reach more than 95.0%,and the performance is stable and reliable.The experimental results can show that the method is indeed feasible and can meet the production requirements...